Power multi-modal spatial intelligent anomaly detection and analysis method and system

Abstract

The present invention proposes a method and system for detecting and analyzing anomalies in electric power multimodal space intelligence, and relates to the technical field of electric power anomaly detection. The method comprises the following steps: constructing and optimizing a three-dimensional model of electric power facilities; based on the optimized three-dimensional model, considering the structural deformation, displacement and environmental factors of the electric power facilities, identifying anomalies of the electric power facilities; calculating the causal relationship strength between each abnormal electric power facility with strong correlation with the three-dimensional space and other electric power facilities in the electric power system; calculating the spatial anomaly degree of each electric power facility with strong correlation with the three-dimensional space; combining the causal relationship strength and the spatial anomaly degree, calculating the total risk of the electric power system. The spatial intelligent three-dimensional electric power anomaly detection and analysis method proposed in the present invention, as a spatial perception module, breaks through the two-dimensional space limitations of the traditional method for anomaly analysis and detection, further expands the depth of anomaly analysis, improves the intelligence level of the electric power system, and promotes the intelligent management and efficient operation of the electric power industry.

Description

Technical Field

[0001] The present invention belongs to the technical field of power anomaly detection, and in particular to a power multi-modal spatial intelligent anomaly detection and analysis method and system. Background Art

[0002] With the rapid development of power systems, the scale of equipment and the operating environment have become increasingly complex. The power industry has gradually entered the era of digitalization and intelligence, involving a wide variety of multimodal data, including images, audio, text, sensor data, and geographic information system (GIS) data. These data not only vary in source, but also have different structures and characteristics, which poses a huge challenge to traditional data processing and analysis methods. In the operation and maintenance of power systems, real-time acquisition and processing of these multimodal data is the key to ensuring the stable and safe operation of the system. However, existing technologies mainly focus on the processing of single-modal data, and it is often difficult to efficiently integrate information from multiple data sources, and it is impossible to fully explore the potential value in various types of data, resulting in low accuracy and efficiency of anomaly detection and system optimization.

[0003] At present, the detection and analysis of abnormal events in power systems still mainly rely on traditional rules and single-modal data analysis methods. Rules are based on fixed thresholds for detection, which are susceptible to changes in the data environment and the increase in equipment complexity, and often have problems of false detection and missed detection. In addition, although single-modal data processing can provide certain information in some scenarios, it is difficult to fully and accurately capture potential abnormal events in the power system because it cannot cross the information gap between different modes. With the continuous complexity of power equipment, networks and environments, existing anomaly detection methods seem to be unable to adapt to increasingly complex data scenarios and dynamically changing operating environments. The complexity of multimodal data poses severe challenges to power systems. There is often a highly complex complementary relationship between modal data. If it cannot be efficiently fused, it may lead to the loss of key information. Although some multimodal data processing technologies have emerged, they often use simple splicing or weighting methods for modal integration, which lacks pertinence and flexibility. This not only leads to low fusion efficiency, but also limits the ability to understand complex multimodal data.

[0004] In addition, the operating environment and equipment conditions of the power system are dynamically changing. Traditional anomaly detection systems often rely on fixed rules and preset models, and lack effective use of real-time feedback information. This static characteristic not only limits the system's ability to adapt to changing environments, but also fails to continuously improve detection and decision-making strategies based on historical data and user feedback. In practical applications, user needs and choices often reflect the personalized needs of anomaly handling, but existing systems rarely incorporate this information into the optimization process. In addition, the heterogeneity of feedback information is also a bottleneck faced by adaptive feedback optimization. Feedback information may come from multiple channels such as user input, device data, external sensors, etc., and its data format, content, and time series characteristics are different. Therefore, how to efficiently integrate the heterogeneous information of user interaction feedback and extract useful feedback optimization signals from it is a difficult problem that needs to be solved urgently.

[0005] Finally, as power equipment and lines become increasingly complex, traditional two-dimensional detection methods can no longer meet the needs of comprehensive understanding and analysis of the spatial relationship of the power system. These methods have extremely limited use of three-dimensional spatial information. This single two-dimensional perspective ignores the structural complexity of power equipment and facilities in three-dimensional physical space. Power facilities such as high-altitude transmission lines and substations with complex terrain are often distributed in high-altitude, low-lying or complex physical environments. Abnormal events in these scenes are often affected by spatial structural deformation, terrain changes, and spatial occlusion. However, traditional methods cannot capture and analyze these key information, resulting in incompleteness and inaccuracy of abnormal detection results. In addition, in the power system, abnormal events such as line shaking and loose infrastructure are usually closely related to environmental factors such as severe wind pressure and heavy rain. Traditional methods are difficult to construct a correlation model between spatial anomalies and environmental factors, and cannot comprehensively evaluate the risk propagation path of spatial anomalies and the chain effects that may be caused. This limitation causes the system to easily miss or underestimate the potential threats when facing anomalies in complex spatial environments. Therefore, there is an urgent need for an intelligent spatial analysis method that integrates multimodal perception information to provide all-round support for power anomaly detection by deeply mining spatial data features and associations. Summary of the invention

[0006] In order to overcome the deficiencies of the above-mentioned prior art, the present invention provides a method and system for intelligent anomaly detection and analysis in electric power multimodal space, constructs a comprehensive electric power anomaly detection and analysis platform, covers spatial analysis, multimodal data fusion and feedback optimization, and forms a complete intelligent anomaly detection framework. A spatial perception module is added, and a spatial intelligent three-dimensional electric power anomaly detection and analysis method for power facilities with strong correlation in three-dimensional space is proposed, breaking through the two-dimensional space limitations of traditional methods for anomaly analysis and detection, further expanding the depth of anomaly analysis, and using multimodal and sensor data to extend the reasoning and analysis of anomaly detection to three-dimensional space. Through this hierarchical and interlocking technical architecture, the present invention effectively improves the intelligence level of the power system and promotes the intelligent management and efficient operation of the power industry.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of the present invention provides a method for detecting and analyzing anomalies in a multi-modal space of electric power.

[0008] The electric power multi-modal spatial intelligent anomaly detection and analysis method comprises the following steps: Obtain point cloud data of power facilities with strong correlation in three-dimensional space, construct a three-dimensional model of the power facilities, and spatially locate the point cloud data; Based on the spatial positioning results, environmental data and GIS data associated with the power facilities are obtained, and the three-dimensional model of the power facilities is optimized using the environmental data, GIS data and point cloud data, wherein the environmental data includes tree growth information, meteorological data and terrain changes; Based on the optimized 3D model, considering the structural deformation, displacement and environmental factors of the power facilities, the spatial intelligence method is used to identify abnormalities of the power facilities; Obtain multimodal data of power facilities with strong correlation in three-dimensional space, and calculate the causal relationship strength between each abnormal power facility with strong correlation in three-dimensional space and other power facilities in the power system based on the abnormal identification results and combined with multimodal data; Based on the spatial coordinates of the three-dimensional spatially strongly correlated power facilities and the spatial coordinates of the environmental factors, the spatial anomaly degree of each three-dimensional spatially strongly correlated power facility is calculated; Considering the propagation impact of each abnormal power facility with strong correlation in three-dimensional space on other power facilities, the total risk of the power system is calculated by combining the causal relationship strength and spatial anomaly degree.

[0009] A second aspect of the present invention provides an electric power multimodal spatial intelligent anomaly detection and analysis system.

[0010] The power multi-modal spatial intelligent anomaly detection and analysis system includes: The three-dimensional model building module is configured to: obtain point cloud data of power facilities with strong correlation in three-dimensional space, build a three-dimensional model of the power facilities, and spatially locate the point cloud data; The three-dimensional model optimization module is configured to: based on the spatial positioning result, obtain environmental data and GIS data associated with the power facility, and optimize the three-dimensional model of the power facility using the environmental data, GIS data and point cloud data, wherein the environmental data includes tree growth information, meteorological data and terrain changes; The anomaly identification module is configured to: based on the optimized three-dimensional model, consider the structural deformation, displacement and environmental factors of the power facility, and use a spatial intelligence method to identify anomalies of the power facility; The causal relationship strength calculation module is configured to: obtain multimodal data of power facilities with strong correlation in three-dimensional space, and calculate the causal relationship strength between each abnormal power facility with strong correlation in three-dimensional space and other power facilities in the power system based on the abnormality recognition result and in combination with the multimodal data; The spatial anomaly calculation module is configured to: calculate the spatial anomaly of each three-dimensional spatially strongly correlated power facility based on the spatial coordinates of the three-dimensional spatially strongly correlated power facility and the spatial coordinates of the environmental factors; The total risk calculation module is configured to consider the propagation impact of each abnormal power facility with strong correlation with the three-dimensional space on other power facilities, and calculate the total risk of the power system by combining the causal relationship strength and spatial anomaly degree. The third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the electric power multimodal spatial intelligent anomaly detection and analysis method as described in the first aspect of the present invention.

[0011] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the electric power multimodal spatial intelligent anomaly detection and analysis method as described in the first aspect of the present invention are implemented.

[0012] One or more of the above technical solutions have the following beneficial effects: The present invention provides a method and system for detecting and analyzing multimodal spatial intelligent anomaly in power, constructs a full-process power anomaly detection platform, and improves the intelligence level of the power system in the fields of equipment management, operation monitoring, etc. from spatial intelligent analysis to multimodal data fusion and feedback optimization, providing strong technical support for the power industry. Through the spatial intelligent three-dimensional anomaly detection and analysis method, the limitations of the traditional two-dimensional detection method are broken through, the spatial perception and anomaly detection capabilities of power facilities are improved, and intuitive three-dimensional visualization decision support is provided. Through the dynamic selection mechanism of hybrid expert model and hierarchical routing, feature extraction and fusion are performed for different modal data to ensure the efficiency and accuracy of data in the processing process, thereby providing accurate input for anomaly detection. Through the multivariate information feedback and iterative optimization mechanism, user feedback, equipment data and external environment information are integrated to form a closed-loop optimization, ensuring that the system can continuously improve the detection strategy according to real-time changes and provide accurate decision support.

[0013] In view of the problem that the prior art lacks effective fusion of multimodal data, the present invention proposes a feature processing and fusion method based on a hierarchical routing strategy. Through the hierarchical routing of the bottom layer, middle layer and top layer, data of different modalities can be gradually processed and fused; wherein, the bottom layer routing is based on a mixed expert model (MoE), and the most suitable expert model is selected to perform preliminary feature extraction on the input multimodal data; the middle layer routing combines the intermediate features of different modal data for further fusion and alignment; the top layer routing comprehensively processes the global features, and finally realizes accurate multimodal feature fusion. The hierarchical optimization strategy ensures the effective flow of data in the entire processing flow, avoids the problem of information loss or incomplete features, and significantly improves the accuracy and reliability of feature fusion.

[0014] The present invention also provides a method for handling power anomalies driven by self-organizing closed-loop feedback and iterative optimization, which introduces a closed-loop mechanism of user selection and feedback, and incorporates the user's selection and feedback in the process of anomaly handling into the system's adaptive learning process. This closed-loop feedback system not only allows users to select the processing suggestions that best meet actual needs from multiple solutions, but also uses the user's selection as the input for the next round of iteration of the system, so that the system gradually adapts to user preferences and personalized needs in the process of self-optimization. By continuously adjusting the generation algorithm and strategy, the system can respond to changes in user needs in real time and provide a personalized solution for power anomaly handling. The significance of this technology is that by combining user feedback with adaptive optimization, the intelligent and personalized handling of power system anomalies is realized, the accuracy and flexibility of anomaly detection and processing are greatly improved, and strong support is provided for realizing more efficient smart grid management and power system operation and maintenance. The proposed multivariate information feedback self-organizing iteration technology breaks through the limitations of traditional anomaly detection and processing methods, enables the system to continuously fine-tune the anomaly handling suggestions according to the latest feedback in each round of iteration, and can automatically adapt to new abnormal situations in complex and changeable power scenarios, greatly improving the system's abnormality handling efficiency and accuracy.

[0015] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0017] Figure 1 This is a flow chart of the method of embodiment 1. DETAILED DESCRIPTION

[0018] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0019] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0020] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0021] Embodiment 1 As mentioned above, in order to solve the problem in the prior art that it is easy to miss or underestimate the potential threats when facing anomalies in complex spatial environments, this embodiment discloses a method for detecting and analyzing anomalies in electric power multimodal space intelligently. By combining spatial intelligence technology with multimodal data fusion methods, the anomaly detection and risk assessment capabilities of power facilities are innovatively improved. By efficiently integrating three-dimensional point cloud data, GIS data, and multimodal perception data, the system can comprehensively analyze the spatial structure, environmental factors, and equipment status of power facilities, identify equipment anomalies, failures, and potential risks in real time, and accurately predict future risks of power facilities. At the same time, through three-dimensional visualization technology, the system intuitively presents risk distribution, equipment anomalies, and prediction results, helping decision makers quickly locate high-risk areas and provide intelligent decision support.

[0022] It can be understood that due to the differences in application scenarios, the power facilities in the power system can be divided into two categories, including three-dimensional space strong correlation power facilities and three-dimensional space weak correlation power facilities. This embodiment performs anomaly detection on the above two types of power facilities. When performing anomaly detection on the two types of power facilities, different measures are taken to improve the accuracy of anomaly detection.

[0023] In general, when performing anomaly detection on power facilities with strong correlation in three-dimensional space: By integrating 3D point cloud data, geographic information system (GIS) data and multimodal perception data, the system can accurately perceive, understand and reason about power equipment and lines in 3D space. The system uses laser radar, drones and other equipment to obtain 3D point cloud data of power facilities, build a detailed 3D model, and combine multimodal sensor data to achieve real-time monitoring and spatial positioning of equipment.

[0024] Furthermore, in three-dimensional space, the system can detect abnormalities such as structural deformation, displacement, and wear of equipment, as well as the impact of environmental factors such as tree growth and terrain changes on power lines. Through spatial intelligence algorithms, the system identifies the type and location of abnormalities, analyzes the spatial relationship between abnormalities and the surrounding environment and other equipment, and uses spatial data mining technology to identify the correlation between abnormalities and possible chain reactions, and assess regional risks.

[0025] Based on spatial data in the time dimension, the system can predict potential risk trends, such as analyzing the growth rate of trees, predicting the risk of line contact in the future, and providing early warning and maintenance suggestions. The system presents the detection and analysis results in a three-dimensional visualization, providing intuitive abnormal distribution maps and risk assessment reports. Users can view abnormal details in a three-dimensional environment to improve the accuracy and efficiency of decision-making. Spatial intelligence enables the system to have the ability to fully perceive and intelligently analyze power anomalies in complex physical environments. Through an in-depth understanding of three-dimensional spatial information, the system provides a higher level of protection for the safe operation of the power system.

[0026] When performing anomaly detection on power facilities with weak correlation in three-dimensional space: A hybrid expert model generalized hierarchical routing multimodal data fusion and knowledge fusion method is proposed. For the multimodal data (such as images, texts, audio, etc.) in the power industry, the system adopts a hybrid expert model (Mixture of Experts, MoE) architecture, and extracts and processes features of different types of modal data through a dynamic expert selection mechanism, and performs deep data knowledge fusion. Each expert model focuses on the deep processing of specific modal data, such as the visual expert model extracts fine-grained features of image data, and the text expert model performs semantic understanding of language data. By dynamically adjusting the selection and combination of expert models, the system can flexibly respond to diverse data inputs and improve the accuracy and efficiency of data and knowledge fusion. In order to further improve the efficiency and accuracy of the fusion of multimodal data and knowledge features, the system introduces a generalizable hierarchical routing mechanism, and intelligently routes multimodal data through a hierarchical decision tree. The hierarchical routing mechanism dynamically selects the most suitable expert model combination to process and fuse data based on the characteristics and contextual information of the input data. The routing strategy continuously optimizes the path of data flow to the expert model through adaptive learning, ensuring efficient and accurate processing in diverse data scenarios. It can optimize the routing strategy according to emerging data types and business needs, and improve the ability to process unseen data.

[0027] In addition, a multi-information feedback autonomous iteration generation technology is proposed. After generating reasoning and decision results, the system collects feedback information through multiple channels, including direct user input, equipment operation data, historical operation records, and external environment changes. Using multimodal feedback fusion technology, the feedback from multiple sources such as text, images, and voice is deeply analyzed, and an adaptive multimodal feedback loop is constructed to re-evaluate and fine-tune the decision after each optimization to ensure the optimal final result. During the iteration process, the system can autonomously identify reasoning paths that are illogical or deviate from expectations and automatically correct them. During the iterative optimization process, the system generates multiple versions of reasoning and decision results, and provides users with choices through version comparison and dynamic presentation. Users can view the changes and improvements between different versions and choose the best solution based on actual conditions. The system can automatically record user choices and feedback, use this information as input for the next round of optimization, and continuously adjust the algorithm and generation strategy through adaptive learning, thereby forming a closed-loop self-improvement mechanism. Finally, the system provides a complete optimized exception handling report and decision plan, which includes exception identification, causal reasoning, handling suggestions, and optimization iteration records, providing users with transparent and reliable decision support.

[0028] Next, the technical solution of this embodiment will be explained in detail.

[0029] like Figure 1 As shown, the electric power multimodal spatial intelligent anomaly detection and analysis method comprises the following steps: Obtain point cloud data of power facilities with strong correlation in three-dimensional space, construct a three-dimensional model of the power facilities, and spatially locate the point cloud data; Based on the spatial positioning results, environmental data and GIS data associated with the power facilities are obtained, and the three-dimensional model of the power facilities is optimized using the environmental data, GIS data and point cloud data, wherein the environmental data includes tree growth information, meteorological data and terrain changes; Based on the optimized 3D model, considering the structural deformation, displacement and environmental factors of the power facilities, the spatial intelligence method is used to identify abnormalities of the power facilities; Obtain multimodal data of power facilities with strong correlation in three-dimensional space, and calculate the causal relationship strength between each abnormal power facility with strong correlation in three-dimensional space and other power facilities in the power system based on the abnormal identification results and combined with multimodal data; Based on the spatial coordinates of the three-dimensional spatially strongly correlated power facilities and the spatial coordinates of the environmental factors, the spatial anomaly degree of each three-dimensional spatially strongly correlated power facility is calculated; Considering the propagation impact of each abnormal power facility with strong correlation in three-dimensional space on other power facilities, the total risk of the power system is calculated by combining the causal relationship strength and spatial anomaly degree.

[0030] The specific implementation steps are: Before building a 3D model of the power facility, it also includes: Determine the power anomaly detection objects, and divide the power anomaly detection objects into three-dimensional space strong correlation power facilities and three-dimensional space weak correlation power facilities based on the application scenario. Including: Analyze the relationship strength between the abnormal detection value of the power abnormal detection object and the three-dimensional spatial data to determine the spatial correlation of the power abnormal detection object; the three-dimensional spatial data includes three-dimensional point cloud data, LiDAR scanning, and geographic information system GIS data; A relationship strength threshold is set. When the spatial correlation is greater than the relationship strength threshold, the power anomaly detection object is judged to be a three-dimensional space strong correlation power facility; when the spatial correlation is less than the relationship strength threshold, the power anomaly detection object is judged to be a three-dimensional space weak correlation power facility.

[0031] It can be understood that the above division method analyzes the relationship strength between the abnormal detection value of the power anomaly detection object and the three-dimensional spatial data and determines the spatial correlation of the power anomaly detection object: The correlation between the anomaly detection value and the three-dimensional point cloud data, LiDAR scanning, and geographic information system GIS data can be calculated respectively, and the three correlations can be weighted and summed to obtain the spatial correlation of the power anomaly detection object.

[0032] The specific determination method is: Power facilities with strong three-dimensional spatial correlation: The spatial correlation is determined by analyzing the strength of the relationship between the facilities and the three-dimensional spatial data. If the abnormal state of the facility is closely related to its spatial location or environmental factors (such as geographical location, trees, wind pressure, etc.), it is classified as a power facility with strong three-dimensional spatial correlation. For these facilities, three-dimensional spatial data (such as three-dimensional point cloud data, LiDAR scanning, geographic information system (GIS) data, etc.) is the key input information for detecting spatial anomalies, position offsets, structural deformations, etc. of the equipment. For example, equipment deformation of transmission lines, spatial interference between trees and power facilities, and interaction between substation equipment and the surrounding environment.

[0033] Power facilities with weak correlation in three-dimensional space: If the abnormal state of the facility mainly depends on the electrical characteristics of the equipment itself (such as load, temperature, fault current, etc.) and has nothing to do with spatial factors, it is classified as a power facility with weak correlation in three-dimensional space. The function and operation of these facilities have a relatively weak relationship with three-dimensional space and are more dependent on the state of the equipment itself, load or other non-spatial factors. For example, the operation of power generation equipment, control centers, certain power equipment, etc. may not be directly affected by spatial factors, but more dependent on the operating parameters of the electrical system, the health status of the equipment, etc.

[0034] Step 1: In the process of 3D modeling and spatial positioning, it is first necessary to process the point cloud data obtained by LiDAR or drones to build a 3D model of the power facility.

[0035] Point cloud data usually contains a large number of three-dimensional coordinate points, which must be filtered, denoised and registered to ensure the accuracy and integrity of the data. The conversion process of point cloud data starts with discrete three-dimensional coordinate points, and these points are connected into patches using algorithms such as Delaunay triangulation or Alpha shape to construct the three-dimensional surface of the power facility. The Delaunay triangulation algorithm processes point cloud data by connecting adjacent points and generating triangular patches, ensuring that the generated mesh has no overlap or gaps, thereby forming a smooth and continuous three-dimensional surface model:

[0036] This process ensures an accurate conversion of the point cloud to the mesh.

[0037] Among them, Pi and Pj are two points in the point cloud, |A| is the area of ​​the triangle they form, and Threshold is the maximum ratio allowed. Through this algorithm, all adjacent points are connected to generate a three-dimensional surface mesh of the power equipment.

[0038] Furthermore, for surface reconstruction of point cloud data, the Poisson surface reconstruction algorithm can be used. This algorithm reconstructs a smooth and accurate three-dimensional surface model by minimizing the errors in the point cloud data. Through reconstruction, the system can not only obtain the geometric information of the device, but also retain the detailed features of the device surface and the surrounding environment, providing a solid data foundation for subsequent anomaly detection and analysis.

[0039] The surface reconstruction formula is as follows:

[0040] in, is the potential function, ρ(x) is the weight value of the point cloud data, represents the Laplace operator. This formula reconstructs the smooth surface of the power facility by minimizing the error of the point cloud data.

[0041] Furthermore, once the 3D modeling is completed, the next step is to spatially locate and transform the coordinates of these point cloud data.

[0042] Since point cloud data may come from different coordinate systems, in order to ensure the coordination of various data sources, all data needs to be converted to a unified coordinate system. A common method is to transform point cloud data through affine transformation. Affine transformation can rotate, scale, and translate coordinates, thereby ensuring that point cloud data from different data sources can be processed in a unified coordinate system. The formula for affine transformation is:

[0043] Among them, Pnew is the converted point coordinate, Pold is the original point coordinate, A is the affine transformation matrix, and T is the translation vector. Affine transformation can realize rotation, scaling, and translation operations to ensure the accurate positioning of point cloud data in the same coordinate system.

[0044] Furthermore, in order to further improve the accuracy of spatial positioning, especially for power facilities in complex environments, the combination of GPS data and lidar data can effectively improve positioning accuracy.

[0045] GPS provides the geographic coordinate information of the equipment (such as longitude, latitude and elevation), while LiDAR provides the three-dimensional relative position data of the equipment. By combining GPS and LiDAR data and applying differential technology, the accuracy of spatial positioning of power facilities can be significantly improved. Differential GPS technology further optimizes the location data of the equipment by calculating the coordinate difference between the measurement point and the reference point. The formula for differential GPS is:

[0046] Among them, ΔPdiff is the differential GPS calculation result, Pmeasured and Preference are the GPS coordinates of the measurement point and reference point respectively.

[0047] Furthermore, errors are inevitable when performing three-dimensional modeling and spatial positioning. Spatial errors may occur due to factors such as sensor errors and positioning deviations. In order to optimize the accuracy of spatial positioning, the system uses optimization techniques such as the least squares method and the Kalman filter algorithm. The least squares method minimizes the sum of squared errors and optimizes the parameters in the model to minimize the positioning error. The optimization formula of the least squares method is:

[0048] in, is the optimized parameter, A is the design matrix, b is the observed value, and the sum of squared errors is minimized to ensure that the optimized three-dimensional positioning error is minimized.

[0049] On the other hand, the Kalman filter dynamically adjusts the filter gain to update the spatial position of the power facility in real time, optimizes the real-time spatial positioning, and ensures that the data is always accurate. The Kalman filter formula is:

[0050] in, is the predicted value at the current moment, zk is the current observed value, Hk is the observation matrix, and Kk is the Kalman gain. This formula optimizes the real-time spatial positioning accuracy of power facilities by dynamically adjusting the Kalman gain.

[0051] Based on the spatial positioning results, environmental data and GIS data associated with the power facilities are obtained, and the three-dimensional model of the power facilities is optimized using the environmental data, GIS data and point cloud data. The environmental data includes tree growth information, meteorological data and terrain changes, specifically: 1. Dynamic environment data collection and preprocessing: By integrating dynamic environmental data from sensors, weather stations, drones and remote sensing equipment, such as wind speed, temperature, humidity, precipitation, etc., the system can obtain real-time environmental changes around power facilities. These environmental data are pre-processed and cleaned to ensure data accuracy and consistency. Through data comparison and trend analysis, the system determines which environmental factors have a greater impact on power facilities, thereby providing accurate data support for subsequent optimization.

[0052] 2. Impact assessment of environmental factors: During the space optimization process, the system analyzes the stress state, fatigue analysis and potential fault points of power facilities based on environmental data. Taking the impact of wind speed on transmission lines as an example, the system will evaluate the stress changes of equipment under different environmental conditions in real time based on factors such as wind speed changes, wind pressure effects and wind direction changes, determine whether the equipment exceeds the safe load range, and then generate an optimized stress analysis model.

[0053] For example:

[0054] 3. Dynamic adjustment and update of spatial model: The system dynamically adjusts the equipment status and environmental impacts in the spatial model based on changing environmental factors. For example, rising temperatures may cause equipment to expand, or tree growth may cause physical interference with transmission lines. The system ensures that the power facility model can reflect changes in the real environment by updating the physical attributes in the model (such as size, location, load, etc.) in real time.

[0055] The key to this dynamic adjustment is to combine real-time environmental monitoring data with three-dimensional spatial data to calculate the coupling effect of environmental changes and equipment status, so as to adaptively modify the facility model. For example, the impact of temperature changes on the metal material of the equipment, or the impact of wind on the line structure, the system can accurately calculate these effects by adding environmental impact factors.

[0056] 4. Optimized model verification and iterative update: After completing the model optimization, the system will generate a preliminary optimized three-dimensional spatial model and compare it with the actual on-site measurement and real-time monitoring data. This step is crucial because it helps the system find possible deviations or errors in the optimization process and further correct the model by comparing the differences between the predicted results and the actual data.

[0057] The comparison process can be achieved through the following formula:

[0058] in, represents the bias of the model, is the optimized model result, It is the actual monitoring data. By comparing the deviations, the system can identify potential problems in the model and make further iterative updates based on the results of field verification.

[0059] 5. Iterative optimization and continuous updating: Based on the comparison and deviation correction of field data, the system will enter the iterative update phase. In this phase, the parameters of the spatial model are continuously updated and its accuracy is optimized by using new monitoring data. After each update, the system will re-evaluate to ensure the accuracy and reliability of the model. Based on the results of each round of updates, the system will optimize the health status prediction, risk assessment and maintenance decision of the equipment, thereby providing continuous support for the long-term stable operation of the power industry.

[0060] Iteration formula:

[0061] in, is the model after the kth round of optimization, is the deviation correction for the kth round, and α is the update coefficient to ensure that the step size of each optimization is appropriate.

[0062] 6. Long-term verification and model application: After multiple rounds of iterative updates, the system will enter a long-term monitoring phase to continuously verify the effectiveness of the optimized model in the daily operation of the power facility. As new environmental data and equipment status data continue to flow in, the system will be further optimized and adjusted based on this data to ensure that the 3D spatial model of the power facility remains efficient, accurate, and reliable in future applications.

[0063] As the impact of environmental factors on power facilities continues to change, spatial model refinement and optimization are necessary steps to ensure model precision and accuracy. By collecting dynamic environmental data related to the equipment (such as wind speed, temperature, etc.), the system can further optimize the three-dimensional spatial model and accurately assess the stress conditions and potential failure risks of power facilities. The system will make modeling adjustments based on these changes to ensure that the power facility model is always consistent with the actual environment and equipment conditions. Finally, after the above steps, the system generates the final three-dimensional spatial model and compares the accuracy of the model through field verification and real-time monitoring data. This verification process is a key step to ensure the reliability and accuracy of the model. Through on-site data comparison, the system can discover potential deviations in the model in real time, and continuously optimize the model through iterative updates to ensure its efficiency and accuracy in future applications.

[0064] Step 2: After the 3D spatial model is built, the system uses spatial intelligence algorithms to monitor the status of power facilities in real time and identify potential abnormalities. Abnormalities usually include structural deformation, displacement, wear and tear of equipment, and the impact of environmental factors (such as the impact of tree growth on lines, terrain changes, etc.) on power facilities.

[0065] Structural deformation detection is achieved by comparing the current state of the power facility with the historical state. Specifically, the system detects structural deformation or displacement based on the deviation of the equipment surface and geometry. If the difference between the current three-dimensional coordinate point of the equipment and the historical coordinate point exceeds the preset threshold, it is judged as abnormal.

[0066] The detection formula for structural deformation is:

[0067] Among them, Pcurrent and Pprevious represent the three-dimensional coordinate points of the device at the current time and the historical time respectively, and ΔP is the deviation of the device structure deformation. When ΔP exceeds the set threshold, the system will mark it as abnormal.

[0068] In addition, displacement detection is performed by comparing the spatial coordinates of adjacent points of the power facility to determine whether the equipment has undergone abnormal displacement. If the displacement exceeds the preset threshold, it may indicate that the equipment has a structural problem. The displacement calculation formula is:

[0069] Where ΔD is the size of the device displacement, and X, Y, and Z are the three-dimensional coordinates of the device. If the displacement value ΔD exceeds the set threshold, the system will mark the device as abnormal.

[0070] Furthermore, the potential impact of environmental factors (such as tree growth, terrain changes, etc.) on power facilities is an important part of power anomaly detection. By combining GIS data with real-time sensor data, the system can evaluate the impact of these factors on power lines and equipment. For example, tree growth may affect the safe distance of power lines, especially in windy or humid environments, where tree growth will put greater pressure on power facilities.

[0071] The system calculates the minimum distance between trees and power facilities. If it finds that the distance is too small, the system will mark the device as a potential anomaly, indicating a possible risk of contact.

[0072] The minimum distance between trees and power lines is calculated as:

[0073] Where d is the minimum distance between trees and power lines, and Rsafe is the set safety distance. If d is less than Rsafe, the system will mark it as a potential contact risk anomaly.

[0074] Furthermore, combined with multimodal data, the causal relationship strength between each abnormal power facility with strong correlation with three-dimensional space and other power facilities in the power system is calculated.

[0075] Specifically, causal reasoning methods are used to deeply explore the relationship between multimodal data and establish a causal chain of abnormal events. Each modal data (such as images, sounds, temperature sensor data, etc.) may provide different abnormal information. The system will conduct a comprehensive analysis of this information to reveal the root cause of the abnormality and possible chain reactions. For example, changes in cracks on the surface of equipment may be related to fluctuations in ambient temperature, or changes in tree growth may interact with changes in line load. By analyzing the correlation between different data sources, causal chain reasoning can deeply identify the potential root causes of abnormal events and reveal potential risks in the system through causal relationships.

[0076] The formula for the strength of causal relationship is:

[0077] in, is the correlation coefficient between power facility i and power facility j, and They represent the observed values ​​of power facility i and power facility j at time t, and is the mean of power facility i and power facility j, and T is the time step. By calculating the correlation between each mode, the system can reveal the potential causal relationship and provide in-depth analysis of abnormal events.

[0078] Furthermore, based on the spatial coordinates of the three-dimensional spatially strongly correlated power facilities and the spatial coordinates of the environmental factors, the spatial abnormality degree of each three-dimensional spatially strongly correlated power facility is calculated.

[0079] Analyze the spatial relationship between abnormal events and the surrounding environment in the three-dimensional space of the power facility, identify the source and potential impact of the abnormality, and further optimize the abnormality detection results. The spatial relationship between power facilities and environmental factors has an important impact on the occurrence and development of abnormal events. For example, the growth of trees, changes in terrain, or position deviations of equipment may all be the cause of a fault. Through spatial relationship analysis, the system can effectively identify the location of the abnormality and its interaction with the surrounding environment.

[0080] The calculation formula of spatial anomaly degree is:

[0081] Among them, Sanomaly (i) is the spatial anomaly of power facility i, which represents the risk level of power facility i, Pi is the spatial coordinate of power facility i, and Ei is the spatial coordinate of environmental factors (such as trees, growth conditions, terrain changes, etc.). , is the weight, and distance(Pi,Ei) is the distance between power facility i and the environmental factor. If the distance distance(Pi,Ei) exceeds the safety threshold, the system marks the device as abnormal. is the structural deformation of the power facility i (such as displacement, cracks, etc.).

[0082]

[0083] , , is the three-dimensional coordinate of device i at the current moment.

[0084] , , is the 3D coordinate of device i at the original time (usually the coordinate when it was initially installed).

[0085] Furthermore, abnormal events in the power system often have a chain reaction, which may lead to further failures in other equipment or areas. For example, equipment failure may cause failure of adjacent equipment through load transfer, thereby affecting the safety of the entire power system. Therefore, the system uses spatial data mining technology to analyze the possible associations between abnormal events and assess the risk of abnormal expansion. The system can not only identify abnormalities in the equipment itself, but also predict the potential impact of abnormal events on other equipment. The chain reaction analysis formula is:

[0086] Among them, Sanomaly(i) is the spatial anomaly of power facility i, λij is the propagation coefficient of power facility i to power facility j, and Cij is the causal relationship strength between power facility i and power facility j, indicating the dependence between the two. Through this formula, the system can calculate the total risk Rtotal of the power system, taking into account the propagation impact and causal relationship between each device.

[0087] When analyzing the total risk of the power system, the system not only calculates the local risk Sanomaly(i) of each device, but also comprehensively evaluates the mutual influence between devices by calculating the propagation coefficient λij and the causal relationship Cij between devices, thereby deriving the overall risk level of the power system.

[0088] Among them, the calculation method of λij can be:

[0089] in:

[0090] βij is the connection strength of the current drive between devices ij:

[0091] in, Indicates the actual current flow between device i and device j, usually obtained through power flow monitoring equipment (such as current sensor). In the power system, this value is a direct indicator of the current transmission intensity between devices. Indicates the maximum current capacity between device i and device j, that is, the maximum current that can be safely transmitted between the pair of devices. It is usually determined by the rated current of the device, the maximum current carrying capacity of the cable, or the design capacity of the device.

[0092] is the physical distance between device i and device j, and the calculation formula is as follows:

[0093] Among them, (Xi, Yi, Zi) and (Xj, Yj, Zj) are the three-dimensional coordinates of device i and device j.

[0094] λij: The propagation effect coefficient of the abnormality of device i on device j. The propagation coefficient is inversely proportional to the connection strength wij and the distance dij between devices. The greater the connection strength, the greater the abnormal propagation effect between devices; the closer the distance, the stronger the propagation effect.

[0095] Step 3: Based on the risk assessment, the system also estimates possible risks in the future through predictive models.

[0096] To this end, the system introduces a prediction model based on time series data, combining historical equipment operation data, environmental change trends and other information to predict future abnormal events. This step uses the Transformer model, an advanced time series prediction method, to predict future risks. Assuming that we have the historical risk data Rtotal(t) of the power system (the total risk Rtotal of the power system at time t), the system can predict the time series through a large Transformer-based time series model:

[0097] Among them, Rtotal(t+k) is the risk prediction value of the power system after k time steps in the future. The time series large model can predict the possible risk trends in the future by learning the historical risk data of the equipment.

[0098] Furthermore, after completing the risk assessment and prediction, the system needs to provide an intuitive three-dimensional visualization to help users understand the risk distribution of power facilities and the abnormal status of equipment.

[0099] Three-dimensional visualization not only helps decision makers clearly view the spatial distribution of equipment failures, but also intuitively presents the predicted risk trends, thereby providing support for decision-making. Based on three-dimensional spatial data (such as LiDAR point cloud data, GIS data, etc.), combined with risk assessment results, the system generates a three-dimensional visualization model of power facilities for users. By combining the abnormality Sanomaly (i) of the equipment and the predicted total risk Rtotal (t) of the power system at time t, the system displays the risk level of each device in three-dimensional space and highlights high-risk equipment by color or marking.

[0100] The system can use the following formula to represent the risk of device i in three-dimensional space:

[0101] in, It is the risk performance of equipment i in 3D visualization, which can be expressed by color depth, mark size, etc. The system dynamically adjusts the risk display of the equipment according to the abnormality of the equipment and the predicted total risk of the power system.

[0102] Furthermore, in addition to displaying the current risk status, the system also displays possible risk events and their trends in the future through 3D visualization. By combining the risk prediction result Rtotal(t) with the 3D model, the system can predict high-risk areas at future time points and issue early warnings to users through dynamic updates. The system can dynamically generate a future risk distribution map based on the future predicted risks (through the prediction formula mentioned above), and update it in the 3D model to display future high-risk areas in real time.

[0103] Next, the technical solution for anomaly detection of weakly correlated power facilities in three-dimensional space will be explained.

[0104] The anomaly detection of weakly correlated power facilities in three-dimensional space includes the following steps: Step 1: Obtain and preprocess the multimodal data of three-dimensional weakly correlated power facilities, perform feature processing and fusion on the preprocessed multimodal data based on a hierarchical routing mechanism and combined with a hybrid expert model, and finally obtain the fused global features; Step 2: Detect power facility anomalies based on the fused global features; Step 3: Use reinforcement learning methods to generate multiple treatment suggestions for power facility anomalies; Step 4: Obtain user feedback on various processing suggestions, input the feedback information into the self-organizing map model for clustering, and dynamically adjust the weight of each modality data in the hybrid expert model based on the clustering results; Step 5: Use the feedback information as a reward for generating multiple processing suggestions in reinforcement learning to optimize the generation of multiple processing suggestions for the next round; Step 6: Iterate the above steps 1 to 5 until the optimal processing suggestion is obtained.

[0105] in: In step one, in order to realize intelligent decision-making in complex scenarios in the power industry, such as equipment fault diagnosis, intelligent monitoring, state prediction, etc., and improve the accuracy of decision-making, it is necessary to process and fuse multimodal data in the power system, such as images, texts, audio, etc., and then use the fusion features after global fusion to make global intelligent decisions. In view of the problems of low efficiency and insufficient precision in multimodal data processing and fusion in the prior art, this embodiment adaptively extracts, processes and fuses multimodal data through a dynamic expert selection mechanism and a hierarchical routing strategy in step one to improve the accuracy and efficiency of data processing.

[0106] The hierarchical routing strategy optimizes the processing path of multimodal data. Through the hierarchical decision tree structure, multimodal data is gradually processed and fused at the bottom, middle and top levels. The bottom-level routing introduces a hybrid expert model and extracts features by screening appropriate expert models. The middle-level routing combines feature alignment between modalities, and the top-level routing comprehensively processes global features to ensure efficient fusion of multimodal data.

[0107] The dynamic expert selection mechanism is implemented based on the Mixture of Experts (MoE) model. The model dynamically selects the expert model that is most suitable for processing the modal data for feature extraction based on the modal characteristics of the input data (such as images, text, audio, etc.). Each expert model performs in-depth processing on data of a specific modality, such as the visual expert model for image data and the text expert model for semantic understanding of text data.

[0108] Furthermore, in the step 1, the pre-processed multimodal data is subjected to feature processing and fusion based on a hierarchical routing mechanism and combined with a hybrid expert model, and finally a fused global feature is obtained, specifically including: Select the most suitable expert model in the hybrid expert model, perform feature extraction on the preprocessed multimodal data, and obtain the underlying routing features; Through the cross-modal attention mechanism, the features of different modalities in the underlying routing features are aligned and fused to obtain the middle-level features of the corresponding modality; The middle-level features of each modality are weightedly fused to obtain the fused global features.

[0109] Specifically, they include: (1) First, multimodal data from the power industry are collected and preprocessed.

[0110] The multimodal data generated by the power system, including images, text, and audio, needs to be standardized to ensure consistency before entering the model.

[0111] For image data, the image is first normalized to the same size and resolution. After preprocessing, the data is uniformly divided into small patches of fixed size (patches), for example, each image is divided into small patches of size 16x16 pixels. These patches are flattened and converted into vector representations as input to VisionTransformer (ViT). Assume that the input image I is divided into N patches, and each patch Ii is linearly projected into a high-dimensional vector space to obtain a sequence of patches: , the embedding of each small block is expressed as , forming the input sequence of ViT , which is used in the subsequent Vision Transformer (ViT) for image feature extraction and analysis.

[0112] For text data, word segmentation and word vectorization are required to convert the original text into a vector representation that can be used for model processing. The purpose of word segmentation is to break down long text into individual words or phrases, while word vectorization represents each word as a high-dimensional embedding vector. For example, given a text sequence , each word Ti is embedded into a high-dimensional vector space , which can be expressed as:

[0113] in, Represents the embedding vector of word Ti. Through this process, the text data is converted into a high-dimensional vector representation that the model can process.

[0114] For audio data, Fourier transform is used to convert the time domain signal s(t) into the frequency domain signal S(f):

[0115] This formula is used to convert audio data from the time domain to the frequency domain so that the spectral features in the audio can be captured by the system and used for subsequent processing. A common processing method is to divide the sound data into short time windows of fixed size, and convert the audio signal of each time window into frequency domain features through Fourier transform. Then, the Fourier transform result is converted into a Mel-spectrogram to better conform to human auditory perception. The spectrum of each time window is further divided into smaller "spectral blocks" (similar to "patches" in images), and then the embedded vectors of the spectrum blocks are further processed by Audio Transformer for feature extraction.

[0116] Through these preprocessing steps, multimodal data is converted into a standardized format that the model can process, laying the foundation for subsequent feature extraction and analysis.

[0117] (2) Afterwards, the data flow path is optimized by introducing a hierarchical routing strategy.

[0118] The hierarchical routing strategy is divided into bottom-level routing, middle-level routing and top-level routing, which correspond to the initial processing of features, the fusion and alignment between modalities, and the global comprehensive decision-making. The core goal of this strategy is to maximize the processing capabilities of each expert model while ensuring the efficient flow and fusion of multimodal data in the processing process.

[0119] The advantage of adopting a hierarchical routing strategy is that it can give full play to the strengths of each expert model and ensure the efficient flow of data in the entire processing chain. The bottom-level routing is responsible for the extraction of basic features, the middle-level routing focuses on the fusion and alignment of different modal features, and the top-level routing makes global decisions and outputs. Through this hierarchical optimization path design, the system can ensure the accuracy of feature fusion and the efficiency of data processing when dealing with complex power scenarios.

[0120] The bottom-level routing stage. After completing the preliminary processing of multimodal features, the bottom-level routing stage is entered. The bottom-level routing strategy assigns the input data to the corresponding primary expert model according to the modality for basic feature processing. For example, image data is processed by Vision Transformer (ViT) in the visual expert to extract global features and complex long-range dependencies in the image. This process ensures that the basic information of each data modality is fully extracted, laying a solid foundation for subsequent processing.

[0121] The feature extraction in the underlying routing is implemented based on a hybrid expert model. The hybrid expert model selects the expert model that best suits the current input data modality type based on a dynamic routing mechanism to further process the features of each modality. For example, image data will be processed by a visual expert, text data by a text expert, and audio data by an expert who specializes in processing audio features. And each modality corresponds to multiple experts, for example, the image modality includes a visual expert No. 1, a visual expert No. 2, and so on.

[0122] Multiple experts within the modality process the data of each modality from different perspectives. For example, the Vision Expert 1 focuses on the extraction of high-level semantic features, such as object detection, facial recognition, or specific scene analysis. This expert may pay more attention to the global information and semantic content in the image. The Vision Expert 2 specializes in processing the spatiotemporal relationship of images, which is suitable for video analysis or time series-related tasks. Similarly, the Text Expert 1 focuses on the understanding of text, such as sentiment analysis of a single sentence, question-and-answer tasks, etc. This expert may focus on processing simple grammatical structures and common expressions. The Text Expert 2 focuses on the deep relationship between words, such as the recognition of synonyms and antonyms, or the understanding of adversarially generated sentences.

[0123] Each expert is not pre-trained for a certain type of data, but is dynamically selected during training based on the input features through a gate mechanism. Different inputs will activate different experts. For some complex tasks, a modality may require multiple experts to participate in processing at the same time. The gate network generates a set of weights for each modality of data, which are used to determine which expert model is most suitable for processing the current input. For example: in the visual modality, for a complex landscape image, multiple experts may be activated to process different levels of image features (semantic features and spatial relationships).

[0124] Specifically, the data of each modality It is assigned to multiple experts of the corresponding modality for processing. The experts directly act on the data of the modality to generate specific expert features. The output of each expert is:

[0125] in, is the processing function of the i-th expert in modality m. The image expert model uses the VisionTransformer (ViT) model, which captures the dependencies between small blocks through a global self-attention mechanism; the text expert model uses a Transformer-based architecture model, which captures the semantic relationship between words in a sentence through a self-attention mechanism; the audio expert model uses the Audio Transformer model, which can capture the long-term dependencies between spectrum blocks by modeling the spectrum blocks of the Mel-spectrogram using a self-attention mechanism. In the process of image and text analysis, the global audio features output by the Audio Transformer can be used for complex audio analysis tasks.

[0126] The underlying routing generates dynamic intra-modality expert weights for each modality’s input data through a gating network. , and use these weights to perform weighted summation on multiple expert outputs of the modality to generate a modality-specific underlying feature representation .

[0127] For a certain modality m (such as text modality T), the system calculates the weight of each expert according to the following formula: :

[0128] in, corresponds to the weight of the i-th expert of modality m; σ corresponds to the activation function (such as Sigmoid); and The weight matrix and bias term corresponding to the i-th expert of modality m, used to learn the rules selected by the expert; Input data features corresponding to mode m.

[0129] The data of each modality will be processed by multiple experts, and the output of the experts will be weighted. Weighted summation to obtain the final underlying routing features :

[0130] in, The underlying routing features corresponding to mode m; The number of experts corresponding to modality m; The i-th expert weight corresponding to modality m; Corresponding to the features of modality m after processing by the i-th expert.

[0131] For each modality (taking text T, image I, and audio A as examples), the underlying routing features are expressed as:

[0132]

[0133]

[0134] These features are based on the data of each modality and incorporate the characteristics of each expert in the modality. The final features are used for modality alignment and feature fusion in the subsequent middle-level routing. The features of each modality are generated independently, and there is no direct interaction between different modalities. This design ensures the independence of modality feature processing and provides pure modality features for subsequent middle-level routing.

[0135] Through the dynamic routing mechanism, each modal data is processed by the most appropriate expert model, and the extracted features can retain their original information to the greatest extent. This dynamic selection and deep feature extraction achieved through the hybrid expert model not only improves the accuracy of feature processing, but also enables the system to flexibly respond to various complex multimodal data inputs.

[0136] Middle-level routing stage. Middle-level routing is responsible for further fusion and alignment of features of different modalities, especially in scenarios where multimodal information needs to be combined, such as combining image features with text features to understand the status or fault information of power equipment. Middle-level routing determines how to fuse by comparing the intermediate features of different modalities.

[0137] Mid-level features The computation of is generated by the cross-modal attention mechanism during the modality alignment process. For a modality m (such as image, text, or audio), It represents the middle-level features generated by the alignment of the modality m in the middle-level routing. It is obtained by aligning attention with features of other modalities. The formula is:

[0138] in, , represented as the query matrix of modality m, which is used to capture the key requirements of modality m; , represented as a bond matrix from all modes, indicating the alignment relationship between modes; , the value matrix from all modalities is used to provide information after feature alignment. WQ, WK, WV are the weight matrices of the attention mechanism, obtained through training.

[0139] Among them, the core of the attention mechanism is based on the query matrix The dot product with the key matrix K calculates the relationship between mode m and other modes, and then uses these relationship weights to perform a weighted summation on the value matrix V:

[0140] in, , K, V are the query matrix of modality m, the key matrix of all modalities, and the value matrix of all modalities, respectively. dk is the dimension of the key, which is used to scale the dot product to stabilize the gradient. Through the above formula, the system can calculate the alignment weights between modality m and other modalities and generate the mid-level features of modality m . Middle level features Focus on the modality and integrate some features of other modalities. This approach ensures that the core features of each modality are retained, while adding supplementary information from other modalities in necessary contexts, thereby achieving information complementarity and enhancement in the modality alignment process. For example, the text modality can obtain more intuitive contextual information through the image modality, while the audio modality can obtain clearer label guidance through the text modality.

[0141] Mid-level features The generation process is through the query matrix of modality m To drive. Query Matrix It reflects the part of the modal feature that needs to be further strengthened, so the calculation result of the intermediate layer feature is centered on the characteristics of mode m. Specifically, the query matrix Designed to be the underlying routing feature from modality m This allows the attention mechanism to focus on selecting information that is useful to this modality from the features of other modalities when calculating, rather than completely integrating all modal features. The main mode is the current mode, but it incorporates supplementary information from other modes through the attention mechanism. This supplementary information comes from the value matrix of other modes. The attention mechanism is based on and key matrix The dot product of is used to calculate the weights, thereby selectively extracting the most relevant information for modality m from other modalities.

[0142] Top-level routing stage. The top-level routing stage is mainly responsible for the comprehensive processing and decision-making of global features. The top-level routing performs global feature aggregation and decision-making by integrating the intermediate features of each modality. For example, when the system processes power equipment failures, it may need to combine the features of the three modalities of image, text, and audio to generate the final fault report.

[0143] The key to feature fusion is that the system should not only retain the unique information of each modality, but also integrate the complementary information between these modalities to maximize the utilization of the overall information. Feature fusion is performed by weighted summation. The features of each modality are assigned different weights. The size of the weight depends on the importance of the modality or the relevance of the task. The weight represents the contribution of each modality to the global decision. Feature fusion can integrate the data features of each modality to generate a unified high-dimensional feature representation. At this point, feature fusion not only retains the uniqueness of a single modality, but also provides richer contextual information through the combination of complementary features, thereby providing support for the subsequent decision-making process.

[0144] In the top-level routing, receive the middle-level features from the middle-level routing , where m represents the type of modality (such as image, text, audio). The top-level routing dynamically learns the weight of each modality in the global task. , weighted fusion of these modal features is performed to generate the final global feature The specific formula is:

[0145] Where M is the total number of modalities (e.g., image, text, and audio modalities, M=3); represents the middle-level features of mode m; is the weight of modality m, indicating the importance of modality m to global feature fusion; weight Can be obtained through learning, satisfying .

[0146] Weight The weight of each modality is dynamically generated through a small neural network or attention mechanism, and the generation formula is:

[0147] Among them, Wα is the parameter of the weight generation network, which can be optimized through training. Therefore, the system uses weights For each modality, the mid-level features Weighted, calculate the weighted features of each mode. Sum the weighted features of all modes to generate the final global feature .

[0148] In this way, the system is able to integrate the features of each modality in a global context and generate a unified output for subsequent tasks, such as fault identification reporting of power equipment, anomaly detection, or intelligent decision-making for other tasks.

[0149] This embodiment introduces a hybrid expert model in feature extraction, and can adaptively select the most appropriate expert model for processing based on the input multimodal data (such as images, text, audio, etc.). This dynamic expert selection mechanism effectively improves the efficiency of multimodal data processing and avoids the redundancy and inefficiency of a single model in traditional methods when processing data of different modalities. Each expert model specializes in processing data of a specific modality, is highly targeted, and can perform deep feature extraction on the data, thereby greatly improving the accuracy and efficiency of feature processing.

[0150] Secondly, the introduction of hierarchical routing strategy greatly optimizes the fusion process of multimodal data. Through the hierarchical routing of the bottom, middle and top layers, the system can gradually process and fuse data of different modalities. The bottom-level routing first selects a suitable expert model for preliminary feature extraction, the middle-level routing combines the intermediate features of different modal data for further fusion and alignment, and the top-level routing comprehensively processes the global features, and finally realizes accurate multimodal feature fusion. This hierarchical optimization strategy ensures the effective flow of data in the entire processing process, avoids the problem of information loss or incomplete features, and significantly improves the accuracy and reliability of feature fusion.

[0151] In addition, the method of this embodiment has extremely high computational efficiency. The hybrid expert model uses a dynamic selection mechanism to only activate the expert model related to the current input mode, avoiding redundant calculations of irrelevant modes, reducing computational overhead, and ensuring that the system can maintain low computing resource consumption when processing large-scale multimodal data. This has important practical value for data scenarios that require real-time processing in the power industry, especially in tasks that require rapid response, such as fault detection and equipment monitoring.

[0152] Finally, the method of this embodiment has strong flexibility and scalability. Whether facing newly added modal types or data processing requirements in different scenarios, the system can adaptively respond to various complex power application scenarios by dynamically adjusting the selection and routing strategies of expert models. This flexibility enables the method to be widely used in various data analysis and intelligent decision-making tasks in power systems, especially in improving the application effects in the fields of power equipment fault diagnosis, intelligent monitoring and state prediction, etc., with significant advantages.

[0153] In step 2 above: After completing the fusion of multimodal information, the system inputs the fused global features into the anomaly detection module, and uses the Trie-based detection agent (TDA) module for real-time analysis to identify potential abnormal events. The TDA module uses a lightweight Trie structure to parse the data step by step in an efficient manner to generate preliminary anomaly handling suggestions. This step converts the detected abnormal conditions into multiple processing solutions, providing diverse options for subsequent user selection and feedback.

[0154] Furthermore, firstly, the TDA module quickly detects anomalies through a lightweight Trie structure. The Trie structure is a prefix tree structure suitable for high-frequency, low-latency data processing, which can efficiently store and retrieve abnormal templates. In the power system, when each piece of data flows in, the TDA module will match it with the nodes in the Trie tree layer by layer, and detect potential abnormal events based on predefined abnormal patterns. The advantage of the Trie structure is that it can dynamically expand nodes to adapt to data updates and changes. In response to the evolving types of anomalies in the power system, the TDA module can add new nodes or adjust existing nodes in real time, thereby achieving flexible anomaly detection. At the same time, the TDA module combines statistical methods such as extreme value theory (EVT) to quantify the detected anomaly frequency and characteristics to ensure the accuracy and timeliness of anomaly detection.

[0155] Furthermore, the anomaly detection based on the Trie structure determines whether an anomaly occurs by counting the occurrence frequency of log templates. Assume that the log template set is ,in, Represents each log template respectively; the number of occurrences of the i-th log template is N(si), and for each template si, count its number of occurrences N(si) in the time window. Normalized to get the occurrence probability:

[0156] It can be understood that before counting the occurrence frequency of the log template, the comprehensive features are matched layer by layer with the nodes in the pre-built Trie tree, and the occurrence frequency of the log template corresponding to the node after matching is counted, specifically: Extract the feature vector Ffinal from the current device state, and calculate the cosine similarity between the feature vector Ffinal and the feature vector FTrie of each path in the Trie tree. The formula is as follows:

[0157] Where Ffinal is the feature vector of the multi-modal comprehensive features after the hybrid expert model is fused, and FTrie is the feature vector of the Trie tree path; When the calculated similarity is greater than the set threshold, the log template is considered to be matched successfully; For each successfully matched log template si, update its occurrence frequency N(si).

[0158] Furthermore, the above contents are further elaborated as follows: (1) Definition of the above log template: A log template is a structured pattern extracted from the original log information, usually consisting of a fixed part and a variable part. For example, the template for the log message "Task 123 completed in 2ms" may be "Task <id>completed in <time>". In the Trie tree, the content stored in the node is the segment (or key field) of the log template. By inserting the log segment into the Trie tree, each log will correspond to a path in the tree layer by layer.

[0159] (2) Related description: The nodes of the Trie tree represent the components of the log template (such as field values, paragraph contents).

[0160] The log template as a whole corresponds to the path in the Trie tree, and the path from the root node to the leaf node forms the complete structure of the log template.

[0161] (3) Log template setting location: Log templates are usually automatically generated by log parsing tools (such as Drain or LogParser) at the beginning of system operation, or they can be defined by experts and stored in the knowledge base. The setting location and purpose are as follows: (4) Storage location: Log templates can be stored in the database or memory of the log management module as input for Trie tree construction.

[0162] (5) Relationship with Trie tree: Templates serve as the source of Trie tree node data, and the collection of all templates forms the structure of the Trie tree for efficient matching and statistics.

[0163] (6) Log template frequency statistics method: Statistical mechanism: The statistics of log frequency are implemented based on the layer-by-layer matching of the Trie tree. The specific steps are as follows: Log parsing: Each original log is decomposed into fields or segments according to template rules.

[0164] Trie tree matching: Match the decomposed log segments to the nodes of the Trie tree layer by layer.

[0165] Count update: After a successful match, the counter of the number of occurrences is increased at the leaf node (i.e. the complete template position).

[0166] (7) Algorithm flow: Initialize the Trie tree: insert the log template set into the Trie tree, where each node corresponds to a segment.

[0167] Real-time statistics: find the corresponding leaf nodes through layer-by-layer matching and count them incrementally.

[0168] Dynamic update: If the log template set changes (such as adding a new template), it can be achieved by inserting a new path in the Trie tree.

[0169] To identify anomalies, the Extreme Value Theory (EVT) is used to calculate the anomaly score for each template. is a random variable representing the number of times the template appears, and its extreme value distribution can be expressed by the generalized extreme value distribution (GEV):

[0170] Among them, ξ, μ, σ are shape parameters, location parameters and scale parameters.

[0171] The score calculated based on extreme value theory can help determine whether a log template is abnormal. Specifically, when the probability of a log template occurring is When a certain threshold is exceeded, the system marks it as abnormal.

[0172] Furthermore, the above contents are further elaborated as follows: (1) Extreme Value Theory (EVT): Extreme value theory studies the extreme behavior of data (such as maximum or minimum values). In log anomaly detection, extreme value theory can be used to evaluate whether the extreme frequency of occurrence of log templates deviates from the normal distribution.

[0173] The generalized extreme value distribution (GEV) is used to model the extreme behavior of random variables (such as the number of template occurrences), and the shape parameter ξ, location parameter μ, and scale parameter σ describe the characteristics of the distribution.

[0174] (2) Probability of occurrence of log template P(si): P(si) is the frequency ratio of the occurrence of log template si, which is a relatively simple statistic used to measure the prevalence of the template in log data.

[0175] (3) Key points of the relationship between extreme value theory and P(si): The connection between extreme value theory and P(si) is established through the number of template occurrences N(si). The details are as follows: Number of occurrences N(si): N(si) is the actual count value of log template si within a certain time window and is the basis of frequency statistics.

[0176] Calculation of probability P(si): is a normalized statistic Extreme Value Theory Modeling: Treat N(si) as a random variable and analyze its extreme value distribution (such as the maximum number of occurrences) to identify abnormalities. EVT uses the GEV distribution to fit the extreme cases of N(si) and determines whether it is abnormal by setting a threshold.

[0177] (4) The relationship between extreme value theory and P(si): Extreme value theory provides a description of the tail behavior of probability distribution, namely, the extreme value distribution model of N(si), focusing on those situations whose occurrence frequency is far higher (or lower) than the average level.

[0178] P(si) is a general probability estimate for each template, which can quickly determine the universality of the template.

[0179] (5) Combined use: In this embodiment, P(si) is used to preliminarily filter abnormal templates (eg, select templates that may appear more frequently).

[0180] EVT applies more detailed tail distribution modeling to the filtered templates to confirm whether they are anomalous.

[0181] (6) The process of calculating anomaly scores based on EVT: Log template frequency statistics: For each template si, count its occurrence times N(si) in the time window. Normalize to get the occurrence probability:

[0182] Extreme value distribution fitting: Use the historical data of N(si) to fit the generalized extreme value distribution GEV(ξ,μ,σ).

[0183] The parameters ξ, μ, σ are calculated by maximum likelihood estimation (MLE) or other methods.

[0184] Calculate the anomaly score: For the number of occurrences N(si) of the new log template, calculate its cumulative distribution function value FGEV(N(si)) in the fitted distribution.

[0185] Determine whether it is abnormal based on the set threshold (such as 95% or 99%).

[0186] (7) Combine extreme value theory with P(si) to detect anomalies: If P(si) is lower than a certain probability threshold, it is directly marked as an anomaly (indicating that the template itself is rare).

[0187] If P(si) is high, but EVT determines that FGEV(N(si)) is at the tail, it can also be marked as abnormal (indicating that frequent occurrence may represent system abnormality).

[0188] The extreme value anomaly score Si = 1 − FGEV(N(si);ξ,μ,σ), where FGEV is the cumulative distribution function of the generalized extreme value distribution.

[0189] As an example: Assume that the historical data of the number of occurrences of log template s1 is [50, 55, 60, 58], and the number of occurrences in the current time window N(s1)=120: Very high, exceeding the preset threshold, requiring further analysis.

[0190] Using historical data to fit the GEV distribution to N(s1), we get ξ=0.1, μ=58, σ=5.

[0191] Calculate the cumulative distribution value FGEV(120)≈0.99 for the current N(s1)=120, and the anomaly score S1=1−FGEV(120)=0.01.

[0192] If S1 exceeds the threshold (such as S1>0.95), S1 is judged to be abnormal.

[0193] Furthermore, after detecting an anomaly, the system generates initial processing suggestions and proposes multiple solutions for users to choose from. Based on the detection results, the TDA module analyzes the severity and possible impact of each anomaly, and then generates multiple versions of processing suggestions. Each recommended solution contains different operating steps, risk assessments, and response measures to better adapt to diverse scenarios. For example, for the same abnormal event, the system may generate three processing solutions: one is a low-risk and low-cost solution that is suitable for rapid response; another is a high-security solution that includes more comprehensive inspections and detailed processing steps; and another is an emergency solution that focuses on rapid repairs to avoid greater losses. By providing multiple processing solutions, the system not only increases the flexibility of user selection, but also improves its adaptability to the needs of different industries.

[0194] In step three of this embodiment, a reinforcement learning method is used to generate multiple processing suggestions for power facility anomalies: In the power system anomaly handling scenario, reinforcement learning gradually optimizes strategies from system status, user feedback and execution results through interactive learning. The key to generating multiple handling suggestions is: Mapping anomaly detection results to the state of reinforcement learning; For each exception, define an action space corresponding to different processing solutions; The reward mechanism is used to evaluate the effectiveness of suggestions and optimize the generation strategy.

[0195] (1) Define the reinforcement learning problem State: The state is defined as the current abnormal state of the power facility, which may include the abnormality type, severity, impact range, etc.

[0196] The design of the state vector S can combine the output of the anomaly detection module (such as the detection results of the TDA module) and historical data.

[0197] Action: Actions are different processing solutions generated by the system, for example: Quick fix solutions (low risk, low cost); Comprehensive inspection plan (high assurance, detailed steps); Emergency response plan (prioritize system recovery).

[0198] The action space can be extended from predefined templates to dynamically generated ones, combining the operating specifications and historical experience of the power industry.

[0199] Reward: Rewards measure the actual effectiveness of the proposed solution, including: user feedback (such as user satisfaction score), system performance indicators (such as processing completion time, residual risk assessment), and cost and resource consumption (such as operation cost and time).

[0200] The reward function can be designed as: ; in, is the weight parameter.

[0201] (2) Training reinforcement learning model: Modeling method: Use the Deep Q Network (DQN) or Policy Gradient method to train the model. Input state S and output the Q value or probability distribution corresponding to each solution.

[0202] Training process: Initialization: construct an initial strategy and generate solutions in a random or rule-based manner; Interactive learning: The system selects actions (processing suggestions) based on the current state S, observes user feedback or execution results, and records rewards R; Policy update: Optimize the policy function based on the reward so that the recommendations generated in the future are more in line with the goal.

[0203] Generates various treatment suggestions: Diversified processing solution generation: In the action space of reinforcement learning, multiple solution templates are predefined for each anomaly: Quick Fix: Actions are aimed at saving time and resources; Comprehensive inspection: Actions cover more complex diagnostic and repair steps; Emergency response: Actions prioritize control and mitigation of potential hazards.

[0204] The model dynamically adjusts the action generation rules by learning user preferences and actual execution results.

[0205] Introducing stochasticity or multi-objective optimization: To ensure diversity, an exploration mechanism can be introduced in action selection (e.g. Strategy). Multi-objective optimization (Pareto optimization) is used to generate recommended combinations that meet different needs.

[0206] (3) Closed-loop optimization of user feedback: After the user selects a solution, the system records the user's choice and execution effect, and uses it as feedback for the next round of optimization. The reward function and action generation strategy of the reinforcement learning model are adjusted based on the feedback, so that future processing suggestions are more in line with actual needs.

[0207] (4) Example description: Assume that an overload exception occurs in a certain device, and the system status S includes: exception type: overload. Severity: high. Impact range: medium.

[0208] Three processing solutions generated by reinforcement learning: 1. Quick fix: Action: Reduce current load and redistribute power.

[0209] Rewards: Low cost, but some risk may remain.

[0210] 2. Comprehensive inspection plan: Action: Shut down the equipment, fully check the load and operating parameters, and perform equipment calibration.

[0211] Rewards: High cost, but risk is largely eliminated.

[0212] 3. Emergency treatment plan: Action: Start the backup equipment and switch the power supply immediately.

[0213] Reward: Power restored quickly, but further follow-up is required.

[0214] Example formula: In state S, the reinforcement learning model selects action Ai:

[0215] Select the optimal solution based on the Q value, or generate multiple optimal solutions at the same time.

[0216] Furthermore, finally, the system pushes the generated processing scheme to the user, waiting for the user's selection and feedback. The user can view the detailed information of each processing scheme through the interface, including potential risks, expected effects, implementation steps, etc. The user selects the appropriate scheme according to their own business needs and actual conditions, and can provide feedback. This feedback will be used as input for subsequent iterative optimization, helping the system to gradually adjust the generation strategy and further improve the quality of anomaly detection and processing suggestions. Through this flexible and diverse scheme generation and selection mechanism, the system can meet the needs of different scenarios in the power system with higher accuracy and personalization, laying a solid foundation for subsequent adaptive learning and closed-loop optimization.

[0217] After generating a preliminary anomaly handling solution, the system will present multiple solutions to the user for selection. The user can view the specific details of each solution through the interface, make a choice based on actual needs, and provide feedback on the solution. User feedback will include satisfaction with the solution, desired improvement directions, or personalized needs. The system will record and store this feedback information to provide data support for subsequent iterative optimization and adaptive learning, thereby continuously improving the accuracy and personalization of anomaly detection and handling suggestions.

[0218] Furthermore, first of all, the system displays a variety of processing solutions for users to choose from. On the user interface, the system presents the details of each solution in a clear and actionable format, including: specific operation steps, expected results, risk assessment, required resources and time, etc. Users can compare the pros and cons of different solutions at a glance. For example, some solutions may focus on quick fixes to reduce system risks, while others may focus on thorough repairs of the root causes, but they take a long time. By providing a wealth of choices, the system can better meet the needs of users in different scenarios and give users the initiative in the exception handling process.

[0219] Furthermore, users can submit feedback after viewing the solutions. User feedback contains many aspects of information: for example, users can give a satisfaction evaluation of the implementation effect of each solution, indicate whether the solution meets actual needs, or make personalized suggestions for improvement. In addition, users can also suggest new solutions or adjust certain steps in existing solutions to better meet specific operational requirements. The types of feedback information may include: operational difficulty, resource consumption, expected effects and risks, etc. The user's actual experience and feedback information can play an important optimization role in this system with diverse choices, helping the system to better understand and adapt to user needs.

[0220] Furthermore, finally, the system uses the user's choice and feedback as input for iterative optimization. The system will record the feedback information of each solution, store it in the feedback database, and use it as the basic data for adaptive learning in the subsequent solution generation and optimization process. These feedback data can be analyzed by machine learning methods. For example, the system can use reinforcement learning algorithms to use user choices and satisfaction as reward signals to gradually optimize the generation strategy and solution recommendation logic. By continuously analyzing and learning user feedback, the system can gradually adjust and optimize the anomaly detection solution to make it more intelligent and personalized, thereby generating processing suggestions that are more in line with the actual needs of users in subsequent iterations. Through this closed-loop feedback mechanism, the system realizes adaptive improvement of the anomaly handling process, significantly improving the efficiency and accuracy of anomaly detection and processing in the power system.

[0221] In step 4, feedback information from users on various processing suggestions is obtained, the feedback information is input into the self-organizing map model for clustering, and the weight of each modality data in the hybrid expert model is dynamically adjusted based on the clustering results: In this embodiment, based on user selection and feedback, the system builds a dynamic feedback loop through a self-organizing algorithm (such as a self-organizing map SOM) to continuously optimize the generation process of exception handling suggestions. This step involves further fusion adjustment of multimodal information, dynamic optimization of the feedback update mechanism, and iterative improvement of exception handling suggestions. Through this closed-loop optimization system, the system can continuously adjust the anomaly detection model parameters and generation strategy to better meet user needs and ensure the efficiency and accuracy of the processing suggestions.

[0222] Furthermore, after each round of operation, the system collects data from the user's operations, choices, and feedback, including the user's evaluation of the exception handling suggestions (such as satisfaction, accuracy) and the suggested adjustments. For example, for text feedback, large language models such as GPT are used to extract key opinions and semantic features. For rating and evaluation data, their numerical values ​​or label information are used directly.

[0223] Input the collected feedback data into the Self-Organizing Map (SOM) for clustering. SOM can automatically identify patterns in feedback data, find out user preferences for different scenarios and common exception handling requirements, and thus provide a basis for the system to adjust model parameters. Map the feedback data to a low-dimensional space and cluster the feedback samples through SOM. Identify different feedback modes (such as high satisfaction, low satisfaction, specific modal preferences, etc.) and form a distribution map of user preferences. Based on the feedback clustering results, the system can understand which specific scenarios or modalities are more likely to cause user dissatisfaction, thereby locating the optimization direction.

[0224] Furthermore, based on the feedback patterns identified by SOM, the system dynamically adjusts the weights and gating mechanisms of each expert network in the MOE model. For the modalities or preferences of users with high-frequency feedback, the system can appropriately increase the activation weights of the corresponding experts, thereby strengthening the feature extraction and fusion of related modalities.

[0225] Assume that the expert weights of text, image, and audio modalities are , , Each weight value Dynamically adjust according to the results of SOM clustering, and use the adjusted expert weights to generate the final global features , assuming that the calculation formula for the adjusted weight is:

[0226] in: is the importance score of mode m under the feedback mode obtained by SOM cluster analysis; Represents the importance score of the text, image or audio modality, such as the importance score in the case of modality k (when k = text); represents the sum of the importance scores of text, image, and audio modalities; T, I, and A represent text, image, and audio data, respectively, which is equivalent to adding all modalities together (all modalities added together = 1). For example, if the text modality is more important in the feedback, the fT value is larger. γ is a smoothing parameter used to control the difference between the weights of each expert. When the importance of a modality is A higher value will tend to give a higher weight to the mode.

[0227] Furthermore, an iterative optimization of dynamic feedback loop is constructed. In each round of feedback loop, the system performs weighted fusion of multimodal data according to the latest weight setting to generate new exception handling suggestions. After each round of iteration, the system re-collects user feedback and continues to update expert weights and fusion strategies to form a dynamic closed-loop feedback optimization process. Weighted fusion feature representation, in each round of iteration, based on the current updated expert weights , , Generate the top-level routing feature Ffinal.

[0228] In feedback-driven iterative optimization, the system updates the weights based on the previous round of user feedback. , , This process is repeated until the fusion features meet the user's needs and achieve the best performance.

[0229] In addition, you can set the convergence condition when the expert weight , , When the change of is less than the threshold ϵ in several consecutive rounds or the fusion feature performance is stable, the iteration is stopped.

[0230] Furthermore, the system continuously optimizes the anomaly detection algorithm based on new anomaly types. Based on user feedback information, the system dynamically adjusts the parameters and thresholds of the anomaly detection algorithm, for example, using extreme value theory (EVT). Specifically, the system re-evaluates the anomaly thresholds and category boundaries in the detection algorithm in each iteration to more accurately identify new anomalies. For example, if the system detects that certain anomaly categories appear frequently in user feedback, the system will pay more attention to these categories in the updated iteration and appropriately lower their detection thresholds to enhance the sensitivity of the algorithm.

[0231] Furthermore, after completing the multimodal optimization and feedback update, the system regenerates and iterates the exception handling suggestions. The processing suggestions after each round of optimization are evaluated and fine-tuned based on the latest feedback data to ensure the applicability and rationality of the suggestions. When generating new processing solutions, the system considers previous user feedback as key input to adjust the generated recommendation content and recommendation logic. These suggestions are designed to adapt to dynamically changing power system scenarios, such as abnormal conditions that arise in new environments or under new conditions. After outputting the new processing solution, the system continues to collect user choices and feedback in order to enter the next round of feedback optimization. This feedback-driven iterative optimization process not only improves the accuracy of anomaly detection and processing, but also gives the system the ability to gradually improve itself, realizing the intelligent, adaptive and personalized power anomaly processing.

[0232] In step 5, the feedback information is used as a reward for generating multiple processing suggestions in reinforcement learning to optimize the generation of multiple processing suggestions for the next round: Based on anomaly detection and user feedback, the system further adopts adaptive learning and closed-loop optimization mechanisms, converting user feedback into reward signals through reinforcement learning (such as deep Q-learning algorithm), and continuously adjusting and optimizing the generation strategy of the model. With the continuous input of user feedback, the adaptive learning module dynamically adjusts the generation parameters so that the anomaly handling solution gradually meets the actual operation requirements and user preferences. Through this closed-loop system, the system can identify and adapt to changes in user needs, realize the personalization and accuracy of anomaly handling suggestions, and gradually accumulate experience to form a self-learning cycle.

[0233] Furthermore, first, the system incorporates user feedback into the reward mechanism of reinforcement learning. After each generation and processing of suggestions and obtaining user feedback, the system uses the user's choice and feedback as a reward or penalty signal to update the strategy of the generation model. Specifically, the Q-learning algorithm uses user feedback to adjust the Q value, and uses the learned rewards (such as high satisfaction feedback) or penalties (such as low satisfaction feedback) to guide the strategy adjustment when generating solutions next time. The system continuously accumulates user feedback data during the generation of multiple versions of suggestions, thereby adjusting the strategy and maximizing the long-term accumulated rewards. This method can not only optimize solution generation in the current scenario, but also accumulate experience points to provide support for future solution generation.

[0234] Furthermore, the system optimizes strategies through the deep Q-learning algorithm to ensure adaptability. Deep Q-learning combines the advantages of Q-learning and deep neural networks and can handle complex features and state spaces. In the deep Q-learning architecture, the system inputs the feature vectors of multimodal data into the deep neural network model together with user feedback, and uses the neural network to predict the reward values ​​of different schemes. As the number of iterations increases, the system continuously updates the network weights so that the recommendation generation strategy in a given power anomaly scenario is more in line with actual needs. Since deep Q-learning can handle complex feature relationships in high-dimensional state space, the system can learn user preferences and scenario changes more comprehensively through this method.

[0235] In order to continuously optimize the generated exception handling suggestions through user feedback, the system introduces a reinforcement learning model. Suppose at time step t, the suggestions generated by the system The reward is , the status is , the goal is to learn a policy , so that the cumulative reward maximize:

[0236] Where γ is the discount factor; k represents the delay time of the reward rt+k relative to the current time step t. If k=0, it means that the reward rt+k=rt is the reward obtained immediately at the current time step t. If k>0, it means that the reward rt+kr is the delayed reward obtained at the future time step t+k. It represents the immediate reward obtained by the system after executing the action according to the current strategy π at time step t+k.

[0237] When reinforcement learning algorithms such as Q-learning update their strategies, they use the following update formula:

[0238] Among them, α is the learning rate. It represents the expected cumulative reward that can be obtained by selecting action at in the current state st. Q(st,at) is the core function in reinforcement learning, which is used to measure the value of selecting action at in state st. Q(st+1,a) represents the expected cumulative reward that can be obtained by selecting action a in the next state st+1. a is an action in reinforcement learning, that is, the action that the agent can take in a certain state.

[0239] Using user feedback as a reward signal, the system continuously adjusts its strategy so that the generated exception handling suggestions gradually approach the optimal one.

[0240] More specifically, in this embodiment, the following detailed explanations and examples are given in combination with the application scenarios of the power system: (1) Status The specific state Indicates the description and environmental information of the current power system abnormality. It may include the following aspects: 1) Anomaly detection information Exception type: Such as equipment overload, line interruption, overtemperature, communication failure, etc.

[0241] Example value: Device overload.

[0242] Abnormal severity: The impact level of the anomaly, which can be expressed through confidence (such as probability value) or classification label generated by the anomaly detection module.

[0243] Example value: Severe (confidence 0.9).

[0244] Abnormal impact range: The equipment or areas involved, such as the number of substations and line nodes affected.

[0245] Example value: 5 substations are affected.

[0246] 2) Environmental status information Equipment operating parameters: Real-time status information of the equipment, such as voltage, current, temperature, load and other key parameters.

[0247] Example value: The current load rate is 80%.

[0248] Historical processing records: The historical processing data of the current exception in the system, including the solutions used and the effect feedback.

[0249] Example value: Quick fix solutions were used 3 times in the past with a success rate of 80%.

[0250] Power dispatch information: Real-time grid operation information, such as the availability of backup equipment, changes in power demand, etc.

[0251] Example values: A spare unit is available, and the load demand increases by 1.

[0252] (2) Action The specific action Indicates the processing suggestions generated by the system in the current state. It can be multiple predefined solutions or dynamically generated solution strategies.

[0253] 1) Low-risk solution: Quick fixes, suitable for lower risk scenarios, are simple to perform but may have limited effects.

[0254] Example value: Reduce load distribution to alternate lines.

[0255] 2) Comprehensive solution: In-depth inspection and repair of all potential factors that may cause abnormalities, suitable for high assurance requirements.

[0256] Example value: Full shutdown to check equipment operating parameters and calibrate loads.

[0257] 3) Emergency plan: Prioritize quickly restoring power supply or controlling risks, suitable for scenarios where major accidents need to be avoided.

[0258] Example value: Switch to backup power, shut down affected equipment.

[0259] (3) Rewards The specific award Used to measure the effectiveness of the previous round of suggestions, including user feedback and execution results. The reward can be designed as a weighted combination of the following parts: 1) User feedback: The user's satisfaction rating for the generated suggestions, for example, on a scale of 1 to 5.

[0260] Example value: User rating is 4.

[0261] 2) Treatment effect: The degree of reduction in residual risk, for example, the percentage reduction in failure rate.

[0262] Example value: 30% reduction in risk.

[0263] The time it takes to complete a repair, such as how quickly the proposed solution is executed.

[0264] Example value: Completion time is 15 minutes.

[0265] 3) Resource consumption: The cost of implementing the plan, such as manpower, time, power resources, etc.

[0266] Example value: The processing cost is 500 yuan.

[0267] 4) The reward function can be defined as: ; in, The weight is adjustable.

[0268] (4) Application examples of reinforcement learning framework In time : state : The current anomaly is equipment overload, the severity is medium (confidence level 0.85), the impact range is 2 substations, the current load rate is 85%, and backup equipment is available.

[0269] action : The system generates 3 suggestions: Quick fix: Reduce load distribution; Comprehensive inspection: Stop the machine and check all equipment parameters; Emergency treatment: Switch to backup equipment.

[0270] award : The user chooses the quick fix solution, which takes 10 minutes to complete, reduces the risk by 20%, gives a user rating of 5, and costs 300 yuan.

[0271] Reward calculation: .

[0272] Furthermore, based on reinforcement learning, the system generates multiple versions of solutions for users to choose from, in order to gradually optimize the generation strategy. After each round of adaptive learning, the system generates multiple versions of exception handling solutions for users to choose and compare. Each version of the solution covers different risks, costs, and expected effects, and the user's choice will be further fed back to the system to provide data for the next round of learning. Through the selection feedback of multiple versions of the solution, the system can more accurately identify the user's needs in actual operations and gradually adjust the generation strategy. For example, if the user chooses a high-security solution multiple times between a low-cost solution and a high-security solution, the system will be more inclined to recommend similar high-security solutions in future generation, ultimately making the solution generation more in line with user needs.

[0273] Furthermore, the system forms an adaptive learning loop through continuous closed-loop feedback. In the process of generating suggestions, collecting feedback, and learning adjustment strategies, the system gradually accumulates an understanding of user preferences, while adapting to different abnormal situations in the power system. As the adaptive learning cycle continues, the system gradually establishes a self-optimizing closed-loop feedback mechanism, so that the abnormal handling plan can be dynamically adjusted as user needs and operating scenarios change. Through this continuous learning process, the system gradually forms a self-learning ability, which can not only generate the best solution in the current scenario, but also show good adaptability and flexibility in future unknown scenarios, providing strong support for the intelligent and personalized handling of power abnormalities.

[0274] In step 6, the above steps 1 to 5 are iterated until the optimal processing suggestion is obtained: After completing adaptive learning and closed-loop optimization, the system outputs the optimized exception handling suggestions to actual operations, and conducts continuous monitoring and dynamic adjustments. By collecting operational feedback and monitoring the processing effects in real time, the system can continuously update model parameters to ensure that the processing scheme meets the actual scenarios and operating standards of the power system. While continuously collecting and analyzing user feedback, the system maintains the operation of the iterative optimization mechanism, enabling it to adaptively respond to new demands and changing power anomalies. In this embodiment, the criteria for the optimal processing suggestion include: (1) Convergence of cumulative rewards: In reinforcement learning, cumulative rewards are usually an important indicator for evaluating the quality of a strategy. If the cumulative rewards gradually converge after multiple rounds of iterations, it means that the strategy has become optimal, and it can be considered that the optimal processing suggestion has been obtained.

[0275] The change in the cumulative reward is less than a certain threshold (e.g., less than 1% in N consecutive iterations). The average reward of reinforcement learning reaches a predefined value.

[0276] (2) Stability of multimodal data weights: Whether the weight adjustment of the hybrid expert model reaches a stable state can be used as another criterion. When the clustering results and weight distribution in the self-organizing map model change little after multiple rounds of feedback, it means that the system's processing of multimodal data has become stable.

[0277] (3) User satisfaction feedback: If the processing suggestions generated by the system receive high satisfaction feedback from users (such as selecting the same type of suggestions multiple times), it can also be considered that the optimal suggestion has been obtained.

[0278] (4) Iteration limit: In order to control the computational cost, an upper limit on the number of iterations can be set (such as 100 rounds). When the upper limit is reached, the round with the best cumulative reward or user feedback is selected as the final output.

[0279] Furthermore, first, the system directly applies the optimized processing scheme to actual operations. After each abnormal processing scheme is generated, the system pushes the scheme to the operation end, and the user or operator implements it in the actual power system. The system monitors the execution effect of the processing scheme in real time during the operation process, and collects data related to the execution for subsequent analysis. For example, the system can monitor factors such as equipment status, operating parameters, and external environmental changes during the processing process to understand the actual impact of the scheme. These real-time data can help the system evaluate the effectiveness of the processing suggestions, and promptly remind users or automatically generate new suggestions when undesirable situations are found.

[0280] Furthermore, the system continuously adjusts models and strategies based on real-time monitoring and user feedback. During the application of the treatment plan, the system will continuously collect user feedback and adjust model parameters based on the feedback results. This process can be achieved by analyzing information such as user preferences in operation, key indicators in execution results, and satisfaction in feedback. For example, if a treatment plan shows a high failure rate or fails to achieve the expected results in actual operation, the system will identify the shortcomings in the plan and fine-tune the parameters through the adaptive learning module to optimize the generation of future plans. The system adjusts the anomaly detection model based on the latest data feedback to ensure that it is more in line with on-site needs in future operations.

[0281] Furthermore, the system maintains an iterative optimization mechanism that enables it to adapt to new demands and changes in the power system. Over time, the environment and equipment conditions in the power system may change, and the system needs to be continuously updated and adapted. Through the continuous iterative optimization mechanism, the system can dynamically respond to new types of anomalies and user needs, forming a "learning by doing" self-learning mode. This continuous iterative mechanism ensures that the system is always kept up to date and continuously improves the accuracy and effectiveness of anomaly handling. In this way, the system can not only solve the current power anomaly problem, but also quickly adjust and adapt when the power system changes, ensuring that it always maintains efficient operation in a complex and changing power environment.

[0282] Embodiment 2 This embodiment discloses an electric power multi-modal spatial intelligent anomaly detection and analysis system.

[0283] The power multi-modal spatial intelligent anomaly detection and analysis system includes: The three-dimensional model building module is configured to: obtain point cloud data of power facilities with strong correlation in three-dimensional space, build a three-dimensional model of the power facilities, and spatially locate the point cloud data; The three-dimensional model optimization module is configured to: based on the spatial positioning result, obtain environmental data and GIS data associated with the power facility, and optimize the three-dimensional model of the power facility using the environmental data, GIS data and point cloud data, wherein the environmental data includes tree growth information, meteorological data and terrain changes; The anomaly identification module is configured to: based on the optimized three-dimensional model, consider the structural deformation, displacement and environmental factors of the power facility, and use a spatial intelligence method to identify anomalies of the power facility; The causal relationship strength calculation module is configured to: obtain multimodal data of power facilities with strong correlation in three-dimensional space, and calculate the causal relationship strength between each abnormal power facility with strong correlation in three-dimensional space and other power facilities in the power system based on the abnormality recognition result and in combination with the multimodal data; The spatial anomaly calculation module is configured to: calculate the spatial anomaly of each three-dimensional spatially strongly correlated power facility based on the spatial coordinates of the three-dimensional spatially strongly correlated power facility and the spatial coordinates of the environmental factors; The total risk calculation module is configured to consider the propagation impact of each abnormal power facility with strong correlation with the three-dimensional space on other power facilities, and calculate the total risk of the power system by combining the causal relationship strength and spatial anomaly degree. Embodiment 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0284] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the electric power multimodal spatial intelligent anomaly detection and analysis method as described in Example 1 of the present disclosure.

[0285] Embodiment 4 The purpose of this embodiment is to provide an electronic device.

[0286] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the electric power multimodal spatial intelligent anomaly detection and analysis method as described in Example 1 of the present disclosure are implemented.

[0287] The steps involved in the apparatuses of the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0288] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0289] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.< / time> < / id>

Claims

1. The power multi-modal spatial intelligent anomaly detection and analysis method is characterized by: The following steps are involved: Obtain point cloud data of power facilities with strong correlation in three-dimensional space, construct a three-dimensional model of the power facilities, and spatially locate the point cloud data; Based on the spatial positioning results, environmental data and GIS data associated with the power facilities are obtained, and the three-dimensional model of the power facilities is optimized using the environmental data, GIS data and point cloud data, wherein the environmental data includes tree growth information, meteorological data and terrain changes; Based on the optimized 3D model, considering the structural deformation, displacement and environmental factors of the power facilities, the spatial intelligence method is used to identify abnormalities of the power facilities; Obtain multimodal data of power facilities with strong correlation in three-dimensional space, and calculate the causal relationship strength between each abnormal power facility with strong correlation in three-dimensional space and other power facilities in the power system based on the abnormal identification results and combined with multimodal data; Based on the spatial coordinates of the three-dimensional spatially strongly correlated power facilities and the spatial coordinates of the environmental factors, the spatial anomaly degree of each three-dimensional spatially strongly correlated power facility is calculated; Considering the propagation impact of each abnormal power facility with strong correlation in three-dimensional space on other power facilities, the total risk of the power system is calculated by combining the causal relationship strength and spatial anomaly degree.

2. The electric power multi-modal spatial intelligent anomaly detection and analysis method according to claim 1, characterized in that: Before building a 3D model of the power facility, it also includes: Determine the power anomaly detection object. Based on the application scenario, the power anomaly detection object is divided into three-dimensional space strong correlation power facilities and three-dimensional space weak correlation power facilities. The specific determination method is as follows: Analyze the relationship strength between the abnormal detection value of the power abnormal detection object and the three-dimensional spatial data to determine the spatial correlation of the power abnormal detection object; the three-dimensional spatial data includes three-dimensional point cloud data, LiDAR scanning, and geographic information system GIS data; A relationship strength threshold is set. When the spatial correlation is greater than the relationship strength threshold, the power anomaly detection object is judged to be a three-dimensional space strong correlation power facility; when the spatial correlation is less than the relationship strength threshold, the power anomaly detection object is judged to be a three-dimensional space weak correlation power facility.

3. The electric power multi-modal spatial intelligent anomaly detection and analysis method according to claim 1, characterized in that: Obtain point cloud data of power facilities with strong correlation in three-dimensional space and construct a three-dimensional model of power facilities, including: Preprocess the acquired point cloud data, including filtering, noise reduction and registration processing; Use Delaunay triangulation or Alpha shape algorithm to connect point cloud data into patches to construct the 3D surface of the power facility; The three-dimensional surface is reconstructed using the Poisson surface reconstruction algorithm to obtain a smooth three-dimensional model of the power facility; or, Spatial positioning of point cloud data, including: Convert all point cloud data into the same coordinate system; Obtain GPS data of power facilities, combine GPS data with point cloud data and apply differential technology to optimize the location data of power facilities; The least squares method and Kalman filter algorithm are used to optimize the accuracy of power facility location data.

4. The electric power multi-modal spatial intelligent anomaly detection and analysis method according to claim 1, characterized in that: The 3D model of power facilities is optimized using environmental data, GIS data and point cloud data, specifically: Through environmental data, GIS data and point cloud data, the 3D model of power facilities is dynamically adjusted, and the physical properties in the 3D model are updated in real time to ensure that the 3D model can reflect changes in the real environment and obtain a preliminarily optimized 3D model; Acquire actual monitoring data through on-site actual measurement, and verify and iteratively update the initially optimized 3D model based on the actual monitoring data; During iterative updates, the parameters of the 3D model are continuously updated using new monitoring data, and re-evaluated and verified after each update until the 3D model meets the preset accuracy requirements.

5. The electric power multi-modal spatial intelligent anomaly detection and analysis method according to claim 1, characterized in that: Based on the optimized 3D model, considering the structural deformation, displacement and environmental factors of the power facilities, the spatial intelligence method is used to identify abnormalities of the power facilities, including: When identifying abnormal structural deformation of power facilities, the absolute value ΔP of the three-dimensional coordinate point of the same position of the power facility at the current time and the historical time is calculated. When ΔP exceeds the set threshold, it is identified as abnormal structural deformation; When identifying abnormal displacement of power facilities, the spatial coordinate change ΔD between the current and historical moments of adjacent location points of the power facilities is calculated. When ΔD exceeds the set threshold, it is identified as abnormal displacement deformation. When identifying abnormalities in the environmental factors of power facilities, the minimum distance d between trees and power facilities is calculated. When d exceeds the set threshold, it is identified as a touch risk abnormality.

6. The electric power multi-modal spatial intelligent anomaly detection and analysis method according to claim 1, characterized in that: Combined with multimodal data, the causal relationship strength between each abnormal power facility with strong correlation with three-dimensional space and other power facilities in the power system is calculated. The specific calculation method is: ； in, is the correlation coefficient between power facility i and power facility j, and They represent the observed values ​​of power facility i and power facility j at time t, and is the mean of power facility i and power facility j, T is the time step; or, Based on the spatial coordinates of the three-dimensional spatially strongly correlated power facilities and the spatial coordinates of the environmental factors, the spatial anomaly degree of each three-dimensional spatially strongly correlated power facility is calculated. The specific calculation method is: ； Among them, Sanomaly (i) is the spatial anomaly of power facility i, Pi is the spatial coordinate of the power facility, Ei is the spatial coordinate of the environmental factor, , is the weight, distance(Pi,Ei) is the distance between the power facility and the environmental factor; is the structural deformation of power facility i; or, Considering the propagation impact of each abnormal power facility with strong correlation in three-dimensional space on other power facilities, the total risk of the power system is calculated by combining the causal relationship strength and spatial anomaly. The specific calculation method is as follows: ； Where λij is the propagation coefficient of power facility i to power facility j; is the total risk of the power system; or, Also includes: Based on the total risk Rtotal(t) of the power system of power facility i at time t, combined with the prediction model, the risk prediction value of power facility i at time t+k in the future is predicted. ; Visualize the risk of power facility i in three-dimensional space: ； Among them, Vrisk(i) is the risk representation of power facility i in 3D visualization.

7. The electric power multi-modal spatial intelligent anomaly detection and analysis method according to claim 2, characterized in that: The anomaly detection of weakly correlated power facilities in three-dimensional space includes the following steps: Step 1: Obtain and preprocess the multimodal data of three-dimensional weakly correlated power facilities, perform feature processing and fusion on the preprocessed multimodal data based on a hierarchical routing mechanism and combined with a hybrid expert model, and finally obtain the fused global features; Step 2: Detect power facility anomalies based on the fused global features; Step 3: Use reinforcement learning methods to generate multiple treatment suggestions for power facility anomalies; Step 4: Obtain user feedback on various processing suggestions, input the feedback information into the self-organizing map model for clustering, and dynamically adjust the weight of each modality data in the hybrid expert model based on the clustering results; Step 5: Use the feedback information as a reward for generating multiple processing suggestions in reinforcement learning to optimize the generation of multiple processing suggestions for the next round; Step 6: Iterate the above steps 1 to 5 until the optimal processing suggestion is obtained.

8. The electric power multi-modal spatial intelligent anomaly detection and analysis method according to claim 7, characterized in that: The preprocessed multimodal data is processed and fused based on the hierarchical routing mechanism and combined with the hybrid expert model to finally obtain the fused global features, including: Select the most suitable expert model in the hybrid expert model, perform feature extraction on the preprocessed multimodal data, and obtain the underlying routing features; Through the cross-modal attention mechanism, the features of different modalities in the underlying routing features are aligned and fused to obtain the middle-level features of the corresponding modality; The middle-level features of each modality are weighted and fused to obtain the fused global features; or, The multimodal data includes image data, text data, and audio data; the hybrid expert model includes multiple image experts, multiple text experts, and multiple audio experts; a gating mechanism is used to select multiple experts of corresponding modalities for processing the multimodal data; or, The underlying routing features output by the hybrid expert model are obtained by weighted summing the features processed by each expert and the expert weights corresponding to each expert; or, The middle-level features of each modality are obtained through attention calculation. In the attention calculation, the query matrix is ​​calculated based on the features of the corresponding modality, the key matrix is ​​calculated based on the key matrices of all modalities, and the value matrix is ​​calculated based on the value matrices of all modalities. or, An incremental gradient update method is used to adjust the parameters of the hybrid expert model according to new data input.

9. The electric power multi-modal spatial intelligent anomaly detection and analysis method according to claim 7, characterized in that: In the step 2, based on the fused global features, the abnormality of the power facilities is detected, which specifically includes: The fused global features are input into the Trie-based network model, and the fused global features are matched layer by layer with the nodes in the pre-built Trie tree. The occurrence frequency of the log template corresponding to the matched nodes is counted to determine whether an abnormality occurs: When the occurrence probability of the log template is lower than the set detection threshold, it is directly determined that the power facility is abnormal; When the probability of occurrence of the log template is higher than the set detection threshold, the pre-built generalized extreme value distribution GEV model is further combined to determine the abnormality; or, Combine the pre-built generalized extreme value distribution GEV model to determine abnormalities, including: Get the type si of the log template corresponding to the node after matching; Count the actual count value N(si) of the log template si within a certain time window, and use the historical data of N(si) to fit the parameters of the generalized extreme value distribution GEV model; Based on the generalized extreme value distribution GEV model, calculate the cumulative distribution function value of N(si) in the fitting distribution; Based on the cumulative distribution function value of N(si) in the fitted distribution, the extreme value anomaly score is calculated to determine whether the extreme value anomaly score exceeds the set threshold. If so, the power facility is determined to be abnormal; or, Pre-build the nodes in the Trie tree, including: Get the log template and decompose the log template into fields or segments; Match the fields or segments of the decomposed log template to the nodes of the Trie tree layer by layer. Each node corresponds to a field or segment, and the pre-construction of the Trie tree node is completed; or, The fused global features are matched layer by layer with the nodes in the pre-built Trie tree, and the occurrence frequency of the log templates corresponding to the matched nodes is counted, specifically: Extract the feature vector Ffinal from the current device state, and calculate the cosine similarity between the feature vector Ffinal and the feature vector FTrie of each path in the Trie tree. The formula is as follows: ； Among them, Ffinal is the global feature after the hybrid expert model fuses multimodal data, and FTrie is the feature vector of the Trie tree path; When the calculated similarity is greater than the set threshold, the log template is considered to be matched successfully; For each successfully matched log template si, update its occurrence frequency N(si).

10. The electric power multi-modal spatial intelligent anomaly detection and analysis method according to claim 7, characterized in that: In step 3, a reinforcement learning method is used to generate a variety of processing suggestions for power facility anomalies, specifically including: Defining states in reinforcement learning The abnormality detection information and environmental status information of the current power facilities, wherein the abnormality detection information includes the abnormality type, abnormality severity and abnormal impact range, and the environmental status information includes equipment operation parameters, historical processing records and power dispatch information; Defining actions in reinforcement learning A plurality of different treatment suggestions, wherein the plurality of treatment suggestions include different operation steps, risk assessments and countermeasures adapted to various scenarios, and the plurality of treatment suggestions are obtained through a plurality of predefined solutions or dynamically generated solution strategies; Defining Rewards in Reinforcement Learning It is the weighted sum of user feedback, processing effect and resource consumption; Select actions based on the current state, obtain user feedback information and calculate rewards, update the strategy function based on the reward value, and generate multiple processing suggestions; or, The goal of reinforcement learning is to learn a strategy that accumulates rewards maximize: ； Among them, γ is the discount factor; k represents the reward The delay time relative to the current time step t; It represents the immediate reward obtained by the system after executing the action according to the current strategy π at time step t+k; Update the strategy function based on the reward, specifically: ； Among them, α is the learning rate; represents the expected cumulative reward that can be obtained by selecting action at in the current state st; Q(st+1,a) represents the expected cumulative reward that can be obtained by selecting action a in the next state st+1; a is the action in reinforcement learning.

11. The electric power multi-modal spatial intelligent anomaly detection and analysis method according to claim 7, characterized in that: In the fourth step, feedback information from users on various processing suggestions is obtained, the feedback information is input into the self-organizing map model for clustering, and the weight of each modality data in the hybrid expert model is dynamically adjusted based on the clustering result, specifically including: Obtaining user feedback on various processing suggestions, including satisfaction with the solution, desired improvement direction, personalized needs, suggestions for new solutions and certain steps to adjust existing solutions. The types of feedback information include operational difficulty, resource consumption, expected effect and risk; A dynamic feedback loop is constructed through a self-organizing map model, feedback information is input into the self-organizing map model, feedback information is clustered, different feedback modes are identified and a user preference distribution map is formed, wherein the feedback modes include high satisfaction, low satisfaction and preference for a specific mode; Calculate the importance scores of each modality under different feedback modes, adjust the expert weights of each modality data in the hybrid expert model based on the importance scores, and use the adjusted expert weights to generate the final global features , each expert weight Dynamically adjust according to the results of SOM clustering, and the calculation formula of the adjusted expert weight is: ； Where: fm is the importance score of mode m under the set feedback mode; β is the smoothing parameter; Indicates the importance score of the text, image, or audio modality; represents the sum of the importance scores of text, image, and audio modalities; T, I, and A represent text, image, and audio data, respectively; When the weight change of each modality data after adjustment is less than the threshold in several consecutive rounds or the fusion feature performance is stable, the iteration is stopped.

12. The electric power multi-modal spatial intelligent anomaly detection and analysis method according to claim 7, characterized in that: In step 6, the criteria for determining the optimal processing suggestion include: After multiple rounds of iterations, the cumulative rewards of reinforcement learning tend to converge, and the change range is less than the set threshold; or the weight adjustment of the hybrid expert model reaches a stable state; Or the satisfaction score of user feedback reaches the preset target; Or when the preset maximum number of iterations is reached, the processing suggestion generated in the round with the largest cumulative reward is used as the final output.

13. The power multi-modal space intelligent anomaly detection and analysis system is characterized by: include: The three-dimensional model building module is configured to: obtain point cloud data of power facilities with strong correlation in three-dimensional space, build a three-dimensional model of the power facilities, and spatially locate the point cloud data; The three-dimensional model optimization module is configured to: based on the spatial positioning result, obtain environmental data and GIS data associated with the power facility, and optimize the three-dimensional model of the power facility using the environmental data, GIS data and point cloud data, wherein the environmental data includes tree growth information, meteorological data and terrain changes; The anomaly identification module is configured to: based on the optimized three-dimensional model, consider the structural deformation, displacement and environmental factors of the power facility, and use a spatial intelligence method to identify anomalies of the power facility; The causal relationship strength calculation module is configured to: obtain multimodal data of power facilities with strong correlation in three-dimensional space, and calculate the causal relationship strength between each abnormal power facility with strong correlation in three-dimensional space and other power facilities in the power system based on the abnormality recognition result and in combination with the multimodal data; The spatial anomaly calculation module is configured to: calculate the spatial anomaly of each three-dimensional spatially strongly correlated power facility based on the spatial coordinates of the three-dimensional spatially strongly correlated power facility and the spatial coordinates of the environmental factors; The total risk calculation module is configured to consider the propagation impact of each abnormal power facility with strong correlation with the three-dimensional space on other power facilities, and calculate the total risk of the power system by combining the causal relationship strength and spatial anomaly degree.

14. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps in the power multimodal spatial intelligent anomaly detection and analysis method as described in any one of claims 1 to 12 are implemented.

15. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the power multimodal spatial intelligent anomaly detection and analysis method as described in any one of claims 1-12 are implemented.

Images